Development of a cavity-enhanced aerosol albedometer

We report on the development of a cavityenhanced aerosol single-scattering albedometer based on incoherent broadband cavity-enhanced absorption spectroscopy (IBBCEAS) combined with an integrating sphere (IS) for simultaneous in situ measurements of aerosol scattering and extinction coefficients in an exact same sample volume. The cavity-enhanced albedometer employed a blue light-emitting-diode (LED)-based IBBCEAS approach for the measurement of wavelength-resolved aerosol optical extinction over the spectral range of 445–480 nm and an integrating sphere nephelometer coupled to the IBBCEAS setup for the measurement of aerosol scattering. The scattering signal was measured with a single-channel photomultiplier tube (PMT), providing an averaged value over a narrow bandwidth (full-width at half-maximum, FWHM, ∼ 9 nm) in the spectral region of 465–474 nm. A scattering coefficient at a wavelength of 470 nm was deduced as an averaged scattering value over the spectral region of 465–474 nm and used for data analysis and instrumental performance comparison. Performance evaluation of the albedometer was carried out using laboratory-generated particles and ambient aerosol. The scattering and extinction measurements of monodisperse polystyrene latex (PSL) spheres generated in the laboratory proved excellent correlation between two channels of the albedometer. The retrieved refractive index (RI) of the PSL particles from the measured scattering and extinction efficiencies agreed well with the values reported in previously published papers. Aerosol light scattering and extinction coefficients, single-scattering albedo (SSA) and NO 2 concentrations in an ambient sample were directly and simultaneously measured using the albedometer developed. The instrument developed was validated via an intercomparison of the measured aerosol scattering coefficients and NO 2 trace gas concentrations to a TSI 3563 integrating nephelometer and a chemiluminescence detector, respectively.


Introduction
Atmospheric aerosols influence climate by modifying the Earth's energy balance through absorption and scattering of the incoming solar radiation (direct effects), changing the cloud properties and abundance (indirect effects), as well as the thermal structure of the atmosphere and the surface energy budget (semi-direct effects) (Ghan and Schwartz, 2007;Stier et al., 2007).
This radiative forcing (RF) capacity, characterized by the aerosol single-scattering albedo (SSA, ω) and its complex refractive index (RI), is mainly determined by the aerosol optical properties (scattering, absorption and extinction).The evaluation of the aerosol impact on climate thus requires accurate, widespread and unbiased quantification of its optical properties as a function of the solar radiation wavelength, of their chemical composition and size distribution (Boucher et al., 2013).
Development of appropriate and well-adapted measurement technologies for real-time in situ measurement of aerosol optical properties is an important step towards a more accurate and quantitative understanding of the aerosol climate effect (Strawa et al., 2003;Thompson et al., 2008).
Aerosol single-scattering albedo, defined as the ratio of the aerosol scattering (α scat ) to its total extinction (α ext ) Published by Copernicus Publications on behalf of the European Geosciences Union.

W. Zhao et al.: Development of a cavity-enhanced aerosol albedometer
coefficient, governs the relative strength of the aerosol scattering and absorption capacity.The value of SSA ranges from 0 for a completely absorbing particle to 1 for a purely scattering particle (Ramanathan et al., 2001;Chin et al., 2009;Hallquist et al., 2009).The in situ accurate measurement of SSA is a key challenge in atmospheric science and climate change research (Moosmüller et al., 2009;Yu et al., 2012;Petzold et al., 2013).
Since the aerosol extinction coefficient is the sum of the absorption and scattering coefficients, a commonly used method for the determination of SSA is to separately measure two of the three optical parameters -absorption, scattering and extinction coefficients -with different instruments.In general, the aerosol absorption coefficient is measured with filter-based techniques or the photoacoustic spectroscopy (PAS) technique (Sheridan et al., 2005;Slowik et al., 2007;Cross et al., 2010;Lack et al., 2014).The scattering coefficient is usually measured with an integrating nephelometer and the extinction coefficient can be measured with an optical extinction cell or cavity-enhanced/ring-down spectroscopy (Moosmüller et al., 2009).Improving the detection sensitivity and the measurement accuracy for each optical parameter is of the first importance to improve the measurement accuracy of SSA.
Filter-based instruments, such as particle soot absorption photometers (PSAP), aethalometer and multi-angle absorption photometers (MAAP), are simple, low-cost and insensitive to gaseous absorption.These techniques suffer, however, from the fact that the natural suspended state of the aerosol changed after deposition (Subramanian et al., 2007).The measurements are strongly influenced by the filter type, multiple scattering by the filter medium and the angular distribution of the scattered light (Moosmüller et al., 2009).The measurement uncertainties of the filter-based techniques are typically between 20 and 30 % for laboratory-generated, dry, non-absorbing or strongly absorbing particles (Bond et al., 1999).For high relative humidity (RH) or high lightabsorbing organic aerosol loadings, the bias in filter-based light absorption measurement may be larger than 100 %.With real-time correction for scattering artifacts, the MAAP instrument can achieve a measurement uncertainty of ∼ 12 % for pure soot (Cappa et al., 2008;Lack et al., 2008).
The PAS method provides excellent detection sensitivity and time response (0.08 Mm −1 , with 60 s average) for direct in situ measurement of aerosol light absorption.The reported accuracy ranges from 5 to 10 % (Lack et al., 2006;Arnott et al., 2003).Recently, Langridge et al. (2013) reported a laboratory study on aerosol absorption measurement using PAS at high RH.They concluded that the PAS is not a technique well suited to the measurement of aerosol absorption at high RH due to the impact of water evaporation on PAS signal.The recommended RH in PAS measurements should be controlled in the range of 10-30 %.
Regarding scattering measurements with nephelometers, an important limitation is represented by the measurement truncation angles: light scattered at angles smaller and larger than the truncation angles can not be detected.For instance, for TSI 3563 integrating nephelometer, measurements of scattering light are limited to between 7 and 170 • .The truncation errors lead to the underestimation of scattering coefficients, particularly for particles with large size.The uncertainty in scattering measurements using a nephelometer varies from 5 to 50 % depending on the particle size and the relative humidity (Massoli et al., 2009).Theoretical calculations suggest higher truncation errors for absorbing aerosols due to changes in the scattering phase function (Moosmüller and Arnott, 2003).Correction factors for the truncation errors can be calculated using Mie theory based on the knowledge of the measured Ångström exponent, aerosol size distribution and the complex refractive index (RI, m = n + ik, where n and k correspond to the real and imaginary part of the RI, respectively) (Anderson and Ogren, 1998;Massoli et al., 2009;Müller et al., 2009).A nearly ideal integrating nephelometer was developed by Varma et al. (2003).The reported nephelometer used an integrating sphere (IS) coupled to two truncation reduction tubes to integrate the scattered light.The forward (backward) truncation angles were reduced to ∼ 1 Measurements of optical extinction using single pass cells are limited by the detection sensitivity and are of practical use only for laboratory-generated aerosols or near-source aerosol plumes in the ambient atmosphere (Schnaiter et al., 2003;Virkkula et al., 2005;Chartier and Greenslade, 2012).Cavity-enhanced/ring-down spectroscopy provides highly sensitive and accurate methods for α ext measurement.The detection sensitivity can be better than 1 Mm −1 with an accuracy of < 3 % (Sappey et al., 1998;Smith and Atkinson, 2001;Thompson et al., 2002;Brown, 2003;Pettersson et al., 2004;Moosmüller et al., 2005;Kebabian et al., 2007;Abo Riziq et al., 2007;Zhang et al., 2008;Lang-Yona et al., 2009;Massoli et al., 2010;Li et al., 2011;Mellon et al., 2011;Bluvshtein et al., 2012;Michel Flores et al., 2012;Wang et al., 2012).
Separate measurement of the extinction coefficient with the cavity-enhanced/ring-down method and of the absorption coefficient with the photoacoustic technique has been used for highly sensitive measurement of aerosol single-scattering albedo without changing the dispersed state of the aerosol particles (Langridge et al., 2011;Lack et al., 2012).However, as this still involves different instruments for separate measurements of extinction and absorption coefficients under different sampling conditions, it might cause potential errors in the determination of the SSA value because the aerosol optical properties are very sensitive to the sampling conditions such as temperature and RH (Lack et al., 2008).
Various spectroscopic approaches have been developed for simultaneous measurement on an exact same sample volume to overcome this weakness, such as the integrated photoacoustic nephelometer (Abu-Rahmah et al., 2006;Chakrabarty et al., 2007Chakrabarty et al., , 2010;;Lewis et al., 2008;Sharma et al., 2013) and the cavity ring-down nephelometer (Strawa et al., 2003(Strawa et al., , 2006;;Sanford et al., 2008).An aerosol albedometer incorporating a ring-down cavity and an integrating sphere for simultaneous measurement of optical scattering and extinction at a fixed frequency was developed by Thompson et al. (2008) (Dial et al., 2010;Ma and Thompson, 2012;Wei et al., 2013a, b;Ma et al., 2013).The relative measurement uncertainty in SSA achieved by this device, dependent upon the particle loading, is better than 5 % (with detection sensitivities of 2.7 Mm −1 and 0.6 Mm −1 for scattering and extinction, respectively), which holds promise for sensitive measurement of SSA.
In this paper, we report on the development of a cavity-enhanced aerosol single-scattering albedometer based on incoherent broadband cavity-enhanced absorption spectroscopy (IBBCEAS) incorporating an integrating sphere (IS) for direct in situ measurement of aerosol scattering and extinction coefficients in the exact same sample volume.Truncation reduction tubes were used to minimize the truncation angle (reduced to be within ∼ 1.2 • for the forward (backward) truncation angle) as reported by Varma et al. (2003).The cavity-enhanced albedometer employed the IBBCEAS method for the measurement of the aerosol extinction spectrum over the spectral range of 445-480 nm and the scattering signal was measured in an IS associated with a singlechannel PMT (photomultiplier tube), providing an integrated result over a narrow bandwidth of ∼ 9 nm (full-width at halfmaximum, FWHM) in the spectral region of 465-474 nm.A scattering coefficient at a wavelength of 470 nm was deduced as an averaged scattering value over the effective bandwidth and used for data analysis and instrumental-performance comparison.Evaluation of the albedometer was carried out using laboratory-generated particles and ambient aerosol for both scattering and extinction channels.
In the present work, measurement intercomparisons of the cavity-enhanced albedometer developed were carried out using a Thermo 42i NO x analyzer (equipped with a molybdenum converter) for NO 2 trace concentration measurement and a TSI 3563 nephelometer for aerosol scattering coefficient measurement.The good agreement observed in these instrumental intercomparisons demonstrated that the albedometer developed provided a robust method for direct and simultaneous measurement of aerosol scattering and extinction coefficients (and then SSA) and the concentrations of absorbing gas present in the air sample.

Experimental setup
The scheme of the cavity-enhanced albedometer developed in the present work is shown in Fig. 1.The broadband radiation was provided by a blue LED (LedEngin LZ110B200) with an emission spectrum peaked at 465 nm.The LED was mounted on a peltier heat sink to stabilize the emission intensity.Light was coupled directly from the LED into a multimode fiber of 500 µm core diameter with a numerical aperture (NA) of 0.22 (Ocean Optics).The emerging light from the fiber was focused with a 75 mm focal-length achromatic plano-convex lens to the center of a high-finesse optical cavity.A bandpass filter, centered at 450 nm with an FWHM of 40 nm (Thorlabs FB 450-40), was located in front of the cavity.The optical cavity consisted of an integrating sphere and two truncation reduction tubes (200 mm long, with an inner diameter of 18 mm).The beam diameter in the cavity was about 12 mm.Using a well-collimated beam was helpful in reducing the wall scattering effects.High-reflectivity mirrors (LGR, 0.8 in. in diameter and 6 m radius of curvature, R > 99.99 % between 415 and 465 nm) were mounted on each end of the truncation reduction tubes.The distance between two mirrors (d) was 600 mm.Each mirror was isolated from the air sample flow by a purge volume that was continuously flushed with dry zero air at a rate of 0.09 L min −1 to prevent degradation of the mirror reflectivity by aerosol deposition.The distance L from sample inlet to the outlet was about 470 mm.The continuous air sample flow rate through the cavity cell was 1.5 L min −1 at atmospheric pressure (∼ 99 kPa, monitored with a pressure gauge).With this flow rate, the residence time was about 200 s for the present albedometer (with a total volume of ∼ 1.9 L including the truncation reduction tubes).Light transmitted through the cavity was collected with a 50 mm focal length achromatic lens and coupled into a multi-mode optical fiber of 500 µm core diameter and 0.22 NA.The output of the fiber was directly connected to a spectrometer (Ocean Optics QE65000) equipped with a 100 µm wide slit resulting in a spectral resolution of 0.4 nm over the wavelength range of 412-487 nm (measured using a low-pressure mercury lamp emission).Temperature and relative humidity were measured with a hygrometer (Rotronic, model HC2 humidity sensor).
The integrating sphere, machined from solid aluminum, was segmented into two hemispheres with an inner diameter d 0 of 150 mm.Its inside layer consisted of pressed PTFE (with a uniform reflectivity of > 99 % between 200 and 2500 nm).The internal volume of the sphere was about 1.8 L. A 16 mm diameter hole was present at each pole of the hemisphere for the passage of the probe light beam.A third hole of identical size, located on the side wall of one hemisphere, was used for scattering signal measurement using a photomultiplier tube (PMT, ZOLIX PMTH-S1-CR131A).A 20 mm wide light baffle, made of PTFE, was used to prevent light scattered by the medium from directly reaching the PMT.A bandpass filter, which centered at 470 nm with an FWHM of ∼ 9 nm (470 nm +4 −5 ), was located in front of the PMT to eliminate the ambient stray light.The PMT signal was acquired with a data acquisition (DAQ) card (National Instruments, NI PCIe-6351), which provided an integrated scattering signal over the spectral region of 465-474 nm.

Angular nonidealities of the albedometer
The forward-scattering truncation geometry of the cavityenhanced albedometer developed and a plot of truncation angles as a function of the distance d e from the scattering location in the sphere to the exit or entrance aperture are shown in Fig. 2. Following Varma et al.'s (2003) discussion, the forward-scattering truncation angle of our albedometer varied from 3.1 to 90 • with an average of 12.2 • for the integrating sphere without truncation reduction tube.The effective truncation angle α(d e ) = tan −1 [r/(d e + d 0 )] varied from 1.2 to 3.1 • (d e and d 0 are schematized in the figure and r = 8 mm, the radius of the hole presented at each pole of the hemisphere for the passage of the probe light beam.), with an average value of 1.8 • for particles located in the truncation reduction tube at a distance d e from the entrance aperture.
Figure 3 shows the size-dependent truncated fraction of total scattering for various truncation angles (Baynard et al., 2007).tubes and (4) 0-7 • , in the case of TSI 3563 nephelometer (as specified by the manufacturer).The truncated fraction of total scattering was calculated with Mie scattering theory for spherical monodisperse particles with an RI of m = 1.6 + i0 at λ = 470 nm.For 1 µm diameter particle, truncated fractions of total scattering were 0.22 % and 1.4 % with (truncation angle of 1.22 • ) and without (truncation angle of 3.1 • ) truncation reduction tubes, respectively.The truncation reduction tubes compensated for the near-forward-scattered intensity, and reduced the measurement errors in large-particle scattering measurements.This value of 0.22 % was much smaller than the value of 6.4 % from the TSI nephelometer.
The small truncation angle (0-1.22 • ) of our IS system significantly reduced truncation errors for large particles when compared with a TSI nephelometer.

Data retrieval processing
In the IBBCEAS approach, wavelength-resolved aerosol extinction can be calculated using the following equation (Fiedler et al., 2003;Washenfelder et al., 2008Washenfelder et al., , 2013)): where three components included in the measured total extinction α Total Ext (λ)α Aerosol Ext (λ), α Gas Abs (λ) and α Gas Rayleigh (λ) -correspond to the aerosol extinction, gas phase absorption and Rayleigh scattering by the gas, respectively.R L is the ratio of the total cavity cell length (d, the distance between two mirrors) to the real cell length containing the air sample when the cavity mirror is purged with gas flow.R L can be determined using an absorber with known extinction (such as a dilute concentration of NO 2 ) or geometrically measured based on the assumption that aerosols follow the gas flow path and are not present in the purging volumes (Washenfelder et al., 2013).In this work, R L was determined from the absorption measurement of 42 ppbv NO 2 with and without mirror purging.R(λ) is the mirror reflectivity; d is the distance between two cavity mirrors; I 0 (λ) and I (λ) are the light intensities transmitted through the cavity without and with air samples, respectively.In our experiment, both the I 0 (λ) and I (λ) spectra were more conveniently obtained in N 2 or air; the gas Rayleigh scattering was presented in both spectra and hence canceled.The measured extinction can be rewritten as follows (Washenfelder et al., 2013): Broadband extinction measurement with IBBCEAS provides a robust method for simultaneous and selectively quantitative measurement of both aerosol extinction and absorbing trace gases concentrations using a single instrument.The gas phase absorption can be extracted from the total extinction using the following equation: The first term describes the contribution from multiple gas absorptions and the second includes the contribution from wavelength-dependent aerosol extinction.Where n i and σ i are the number density and the absolute absorption cross section of the ith absorber, respectively, s i and t i are the shift and stretch coefficients for each absorber, used to reconstruct an accurate wavelength calibration.The polynomial offset P (λ), varying from linear to fifth order, is used to account for the variation in spectral background, including wavelength-dependent aerosol extinction and spectralbaseline shift (which can be considered as system drift in the extinction measurement).In the present work, a thirdorder polynomial function was used for data retrieval.For a particle-free sample, P (λ) merely represents the spectralbaseline drift including baseline variation due to Rayleigh scattering by air and unspecified background change in spectra resulting from unstable LED emission and/or unstable dark current variation in the CCD (charge coupled device) spectrometer.For this reason, high stability of an IBBCEAS instrument is highly required for high-accuracy measurements of aerosol extinction such that the background drift could be negligible in comparison with the measured aerosol extinction.
Mirror reflectivity R(λ) of the albedometer was determined by introducing gases with different Rayleigh cross sections (Moosmüller et al., 2005;Washenfelder et al., 2008;Zhao et al., 2013;Dong et al., 2013).In this work, the R(λ) was determined from the difference in the transmitted intensities of N 2 and SF 6 .The cavity was flushed with N 2 and SF 6 at 1.5 L min −1 rate for 40 min for each species, until the transmitted light intensity attained a stable value.The Rayleigh cross sections used for the mirror reflectivity calculation were reported by Naus and Ubachs (2000) and Sneep and Ubachs (2005), with an experimental uncertainty in cross section of 1 % for N 2 and 3 % for SF 6 .The mirror reflectivity was found to be about 99.96 % at 470 nm.During the process of mirror reflectivity calibration, the purging zeroair flow was turned off and the cavity was fully filled with The aerosol scattering coefficient, α scat , is proportional to the ratio of the scattering signal (I scat ) measured with a PMT and the transmitted intensity (I trans ) measured with a CCD spectrometer (the same spectrometer used for the IBBCEAS measurement) (Strawa et al., 2003;Thompson et al., 2008): where K and K are the experimentally determined calibration constants that account for the differences in collection efficiency and response of different type of detectors, respectively.When purging gas was continuously introduced into the albedometer, the effective path length and thus the reduction tube length was shortened.However, for particle diameters smaller than 2 µm, the truncation error was smaller than 2 %, and therefore the purging gas effect (R L factor) might be neglected for the scattering measurement of our albedometer.
The calibration of the parameter K can be made based on the assumption of a linear response of the PMT to the scattering light intensity (Anderson et al., 1996).K might be simply calibrated with CO 2 and N 2 scattering processes by the following equation: where α scat_CO 2 and α scat_N 2 are the theoretically calculated Rayleigh scattering coefficients of CO 2 and N 2 .I scat_CO 2 and I scat_N 2 are experimentally measured scattering intensities when the cavity is filled with CO 2 or N 2 , respectively.I trans_CO 2 and I trans_N 2 are the measured transmitted intensity of the cavity (at λ = 470 nm in our case) for CO 2 and N 2 , respectively.
In order to calibrate the scale factor K well, He and SF 6 were used to extend the dynamical range (from 0.3 to 145 Mm −1 ) of the calibration.The Rayleigh scattering cross section for He was fitted to Shardanand and Rao's data (σ Rayleigh He = 1.336 × 10 −17 × λ −4.1287 ) (Shardanand and Rao, 1977;Washenfelder et al., 2013).The cross sections of N 2 , CO 2 and SF 6 were obtained from Naus and Ubachs (2000) and Sneep and Ubachs (2005).Calibration of K was achieved by flushing the cavity with calibration gases and then performing measurements of the I scat / I trans ratio.A linear fit of the theoretical Rayleigh scattering coefficient of each gas to the measured I scat / I trans ratio is shown in Fig. 4a.As can be seen, the measured I scat / I trans signal is linearly correlated with the theoretically calculated Rayleigh scattering coefficient.The intercept of the I scat / I trans ratio was considered as the contribution of the photon counts due to scattering by internal surfaces.
A regression plot of the measured extinction and scattering coefficients for calibration gases is shown in Fig. 4b,  which proves an excellent correlation between the scattering and the extinction measurements (scattering = −0.288(±0.869) + 0.998 (±0.018) × extinction, with R 2 = 0.9987).

Precision and accuracy of the instrument
The detection limits for the measurement of the scattering and extinction coefficients at 470 nm were determined by an Allan variance analysis.Figure 5 shows an Allan deviation plot realized based on 5.5 h time series measurements of a particle-free zero-air sample with a time resolution of 9 s.Longer-term drift of the instrument was observed which was smaller than 2 Mm −1 (as shown in the upper panel of Fig. 5).The scattering measurement channel exhibited the lowest detection limit of 0.07 Mm −1 , with an optimum integration time of 459 s that was much longer than the optimum integration time for the extinction measurement channel (54 s).With 54 s integration time, the detection limits for the scattering and extinction channels were 0.22 Mm −1 and 0.09 Mm −1 , respectively.
In the lower panel of Fig. 5, frequency distributions of the scattering and extinction measurements are shown.A Gaussian distribution was fitted to the histograms to obtain the mean of the zero-air measurements and the standard deviation (Kennedy et al., 2011;Dorn et al., 2013).The 1σ standard deviation of the Gaussian fit is a measure of the instrument measurement precision.The determined extinction measurement precision of 0.51 Mm −1 (in 9 s) is comparable to the result of 0.19 Mm −1 (with 10 s average time) reported by Petzold et al. (2013).A normal distribution was fitted to the histograms.The 1σ standard deviation, SD, is a measure of the instrument precision; "mean" denotes the mean scattering or extinction coefficients.
For aerosol measurement, the accuracy in the extinction measurement is mainly limited by the uncertainties in (1 − R), R L and particle losses in the cavity.The drift of the LED intensity is not included when considering the accuracy of the extinction measurements, since frequent recording of I 0 (for example, every hour) could allow correction for the baseline drift related to the fluctuation in LED emission intensity.The cavity was flushed with N 2 and SF 6 at a rate of 1.5 L min −1 for 40 min for each species, until the transmitted light intensity attained a stable value.Ten different pairs of N 2 and SF 6 transmission spectra under stable conditions were used for mirror reflectivity determination, and then 10 values of the mirror reflectivity were averaged.The mean value used as mean mirror reflectivity and the mean relative error of (1 − R) is less than 1 %.We estimated an uncertainty of 3 % in R L .The particle loss through the system, determined via the measurements from two condensation particle counters installed at the inlet and the outlet, respectively, was estimated to be 2 %.Considering all of the uncertainties, the total uncertainty in the extinction measurement was estimated to be less than 5 %.
The uncertainty in the scattering measurement is mainly caused by the uncertainties in K , the error caused by the angular nonidealities (less than 2 % for particle diameter smaller than 2 µm) and particle loss in the cavity.The uncertainty of K was less than 2 %.The total uncertainty in scattering measurement was estimated to be about 4 %.
The total uncertainty in the measurement of SSA was then estimated to be less than 5 %, where the (1 − R) and R L errors were considered as the total extinction error, while the errors in K , and the angular nonidealities were considered as the total scattering error.Since the scattering and extinction coefficients were measured on the exact same volume, the uncertainty of SSA for monodispersed aerosol due to particle loss could be ignored.However, for particle diameters larger than 2 µm, the influence of truncation errors for the finite acceptance angle measurements may be potential error sources.

Instrument test using laboratory-generated particles
Performance evaluation of the albedometer developed was performed with the measurements of laboratory-generated, monodispersed polystyrene latex (PSL) spheres.The aerosol generation system was the same as used in our previous work (Zhao et al., 2013).Aerosols were generated with a constant output atomizer (TSI-3076).PSL standards of four different diameters (200, 240, 300 and 400 nm) were generated by an electrostatic classifier (TSI differential mobility analyzer, DMA 3080 L) for the evaluation.The particle concentration was determined with two condensation particle counters (a CPC 3775 at the entrance of the cavity and a CPC 3776 at the exit of the cavity).After taking into account the dilution inside the cavity as a result of the purge flow of zero air on the mirrors, the averaged particle number was used for data analysis.
Laboratory-generated NaCl particles were used for the evaluation of particle loss vs. their size (as shown in Fig. 6).The particle loss is determined by the difference in particle concentrations measured by the two CPCs.For particle diameters larger than 300 nm, the particle loss can be ignored.
The time response of the instrument is evaluated using laboratory-generated, monodispersed PSL particles with a diameter of 240 nm. Figure 7 shows the time responses for the measurements of the particle concentration inside the integrating sphere and the measurements of the corresponding scattering and extinction coefficients using the cavityenhanced albedometer.The rise time (from zero to its final stable value) for measurements of particle concentrations varying from 0 to 393 particle cm −3 was about 190 s, and the .6. Laboratory assessment of the particle loss vs. particle size in the developed edometer.rise time for the measurements of the corresponding scattering and extinction coefficients (37 Mm −1 ) was about 206 s.
Figure 8a shows a regression plot of the extinction (α ext ) and scattering (α scat ) coefficients at λ = 470 nm.The scattering and extinction data were averaged over 5 to 10 min sequences after the aerosol number concentration in the albedometer was sufficiently stable.Error bars in the figure correspond to 1σ of the sequence average.For the measurements of different PSL number concentrations or diameters, the cavity was washed with zero air for acquisition of the I 0 (λ) spectrum in order to correct for drifts in the background spectrum.The transmitted and scattered intensities of the particle-free air sample were used to subtract the light scattered by internal surfaces and by gas portions of the sample.The non-absorbing PSL sphere experiments had excellent correlation between the scattering and extinction measurements from the albedometer.
A plot of the experimentally measured scattering and extinction coefficients vs. the averaged value of the measured particle number concentrations (N) is shown in Fig. 8b.The scattering (σ scat = α scat /N) and extinction (σ ext = α ext /N) cross sections for each particle size were obtained by averaging the measurements of α scat and α ext at different concentrations.The scattering (Q Scat = 4σ Scat /πD 2 ) and extinction (Q Ext = 4σ Ext /πD 2 ) efficiencies were obtained as the ratio of the particle cross section to the geometric cross section.A plot of experimental Q scat and Q Ext as a function of particle diameter is shown in Fig. 8c.The retrieval algorithm of the RI was realized by fitting the measured scattering and extinction efficiencies to theoretically calculated values based on a Mie scattering subroutine, reported by Bohren and Huffman for homogeneous spheres (Bohren and Huffman, 1983;Laven, 2006).Best-fit results were obtained by varying the real and imaginary parts of the RI.A set of RI was found by minimizing the "merit function" χ 2 Num −2 , where χ 2 is expressed as (Dinar et al., 2008;Zarzana et al., 2012;Washenfelder et al., 2013) where Num is the number of measurements (of different particle sizes) used in the fit, Q scat,ext (n, k) represents the scattering or extinction efficiencies, and Q is the standard deviation of each measurement of the same particle size but at different concentrations.
The merit function was calculated for a wide range of n and k values, and the value of n and k that gives the lowest χ 2 (χ 2 0 ) was taken for the retrieved RI.The values of n and k that satisfy χ 2 < χ 2 0 + 2.298, which fall within the 1σ error bound of the best measurement (with 68.3 % confidence level of χ 2 distribution), are considered acceptable.Projections of the contour lines (with a contour value of 2.298) on the n and k plane give the standard errors n and k, respectively.
The RI of PSL was retrieved independently with scattering and extinction efficiencies, independently.The retrieved RI was m = 1.676 +0.009 −0.008 + i0.015 +0.009 −0.008 from the scattering channel and m = 1.674 +0.012 −0.012 + i0 +0.003 0 from the extinction channel.Limited by our aerosol generation system, the particles number concentrations were very small for the particle diameters larger than 400 nm.By using the efficiencies measured with small-particle diameters for the fit of the merit function, a non-zero value of the imaginary part of the RI could not be ruled out.Despite a number of previous studies previously performed, the differences between the retrieved RI values still span a range of about 5 % in the visible spectral region which is mainly due to the experimental difficulty in particulate measurements, in particular due to sample-to-sample differences depending on the nature of the preparation (Miles et al., 2010). CE15 As mentioned in the note on Fig. 9, the reference to "Rudich and co-workers" is confusing; Rudich is not first author for the articles cited, and furthermore, Abo Riziq is also listed as an author for both articles (or all three if Bluvshtein et al. ( 2012) is included).Therefore, please simply list the studies here and adapt Fig. 9 (see separate comment on Fig. 9).ALB (Abo Riziq et al., 2007;Lang-Yona et al., 2009;Bluvshtein et al., 2012) found a value of m = 1.597+i0.005at = 532 nm.

Atmos
 CE16 It is unclear whether this reference should actually be in brackets with Abo Riziq and Lang-Yona, and thus whether it part of the "Rudich and co-workers" group of studies, or whether it is listed separately.If it is to be listed separately, please include the word "and" before Bluvshtein; if it is not to be included, please place it within the brackets.
Please see the reply of CE 15.
 CE17 Please confirm or provide an alternative.
Agree with the change.
 CE19 This sentence is problematic.The construction with the verb "interfere" is not correct, but as there is also a noun missing after the adjective "oscillation-like", only a suggestion can be made: "The big oscillation-like ... in the baseline at > 475 nm and i0.0005 at λ = 560 nm.Nikolov and Ivanov (2000) reported a value of m = 1.617 + i0 at λ = 436 nm and m = 1.606 + i0 at λ = 486 nm.Our results of m = 1.676 + i0.015 (retrieved from the scattering channel) and m = 1.674 + i0 (retrieved from the extinction channel) agree with the reported RI values as shown in Fig. 9.This result was a little larger than our previous result m = 1.625 + i0.038, which was probably caused by the large inner volume of the instrument and hence longer residual time (∼ 200 s) and larger conglomeration effects on small-diameter particles.The larger particle loss leads to underestimation of the particle number concentration and overestimation of the extinction and scattering cross sections.Our results were in close agreement with the RI value give by Nikolov and Ivanov (2000) (interpolation of their data gave m = 1.61 + i0 at the wavelength of 470 nm).The difference between our retrieved refractive index and this interpolation value was about 4 %, within the tolerance of the instrumental accuracy (4 % for scattering and 5 % for extinction measurements, 3 % for particle concentration measurement), which confirmed that the calibration method used for determination of the cavity mirror reflectivity R(λ), the scattering parameter K and the parameter R L (determined by calibration, too) was suitable for the aerosol optical-properties measurement.

Ambient measurement
For further evaluation and validation of the instrument developed, field environment measurements were carried out outside the laboratory at the Anhui Institute of Optics and Fine Mechanics (31 • 54 18 N, 117 • 9 42 E) during the period of 18-19 April 2013.Ambient air was sampled through a copper pipe (22 mm inner diameter) with an inlet about 5 m above the ground level.The acquisition time of the albedometer for each data was 9 s (for 1.5 s integrating time per spectrum, and six-spectra averaging).The cavity  was flushed with dry zero air every hour for acquisition of I 0 (λ) spectrum.The transmitted and scattered intensities of a particle-free (and non-absorbing) air sample were used to subtract the light scattered by internal surfaces and by gas portion of the sample.
An example of data retrieval is shown in Fig. 10 for ambient measurement at two different aerosol loadings: aerosol extinction larger than 400 Mm −1 (panels a, b) and lower than 100 Mm −1 (panels c, d) with different fit windows.A full window of 444-481 nm for aerosol extinction determination is shown in panels a and c, and a narrow window of 444-467 nm for NO 2 concentration retrieval is shown in panels b and d.The NO 2 cross section used reference was generated by convolution of high-resolution absorption cross sections reported by Vandaele et al. (2002), with the slit function of the spectrometer at 294 K.The H 2 O absorption cross section was calculated based on the HITRAN 2008 database (HIghresolution TRANsmission molecular absorption database) (Rothman et al., 2009).The large fit error observed around 475-481 nm was due to the low signal-to-noise ratio (SNR) data related to low light transmission from the cavity.The detection sensitivity for ambient air measurements was lower than that obtained in particle-free sample measurement: approximately 6 times lower for the aerosols with extinction larger than 400 Mm −1 and 3 times lower for the extinction smaller than 100 Mm −1 .Under higher aerosol loading conditions (Fig. 10a), the detection sensitivity deteriorated.The big oscillation-like structure in the baseline at > 475 nm (due to the operation of the albedometer on the edge of the cavity bandwidth) and the absorption structure of aerosol around 465 to 470 nm interfered with the NO 2 concentration retrieved from the full window.The polynomial used in the DOAS fit did not completely account for the aerosol absorption feature.The absorption structure was not observed under lower aerosol loading conditions (Fig. 10c).Using an appropriate spectral region, good data retrieval is obtained (as shown in Fig. 10b).
An overview of ambient aerosol scattering, extinction coefficients, single-scattering albedo (SSA) and NO 2 concentration measured by the albedometer developed is shown in Fig. 11.The particle number concentration and the relative humidity are also shown in the upper panel.The relative humidity was measured using the internal relative humidity sensor of the TSI 3563 integrating nephelometer.
NO 2 concentrations retrieved from the IBBCEAS spectra were compared to the values measured with an online NO x analyzer (Thermo 42i).Good agreement between two analytical instruments' measurements can be observed in Fig. 11 (middle panel), except for the period from 21:00 LT on 18 April to 06:00 LT on 19 April, where the results from the NO x analyzer were about 1.2 ppbv larger than the albedometer measurements.This was probably caused by the interference of NO y measured by the NO x analyzer (equipped with a molybdenum converter) (Villena et al., 2012).However, these differences were still around the tolerance of the NO x detection sensitivity (1 ppbv) for the used NO x analyzer.An enlarged drawing of the NO 2 measurement comparison in two selected periods (10:00-15:00 LT on 18 April for high aerosol load conditions and 06:00-14:00 LT on 19 April for low aerosol loading) is shown in Fig. 12. NO 2 concentrations retrieved with different fit windows are also shown in the figure.An appropriate choice of the spectral region with good quality data was very important for accurate data retrieval (Fig. 10b, d).From a correlation plot of 5 min averaged data (Fig. 13), very good agreement was observed between the two instruments for different aerosol loadings (Albedometer = 0.995 × NO x analyzer + 0.465 ppbv, with R 2 = 0.956).The aerosol scattering coefficient measured by the cavityenhanced albedometer developed was compared with the data from an integrating nephelometer (TSI 3563) operating at three wavelengths centered at 453, 554 and 698 nm (the measurements are larger than the values from the nephelometer when the extinction is large.This is due to the fact that under large extinction conditions, large-diameter particles dominated.The truncation error of the TSI 3563 nephelometer caused an underestimation of the scattering coefficient for the nephelometer.In the case of small extinction, fine particles are dominant and their loss due to conglomeration effects was larger in our system (as shown in  those from the TSI 3563 (Albedometer = 1.13 × TSI Nephelometer − 9.44 Mm −1 , with R 2 = 0.994).The slope of 1.13 implicated that the smaller truncation angle of the integrating sphere used in the cavity-enhanced albedometer allowed for the collection of more scattered light compared to the TSI 3563 nephelometer.As shown in Fig. 3, for 1 µm diameter particles, the truncated fraction of total scattering was about 10 % with a truncation angle of 7 • .And this value was increased to 20 % for particles of diameter of 1.5 µm.The intercomparison between the albedometer and the TSI nephelometer demonstrated the performance of our instrument for ambient air measurement.

Conclusions
The cavity-enhanced methods require very stable light sources.The LED is a promising new type of light source, with long lifetime and low energy consumption and it is more compact than commonly used broadband arc lamps (Ball et al., 2004).High-quality diode laser current and temperature controllers are usually used as LED controllers.In this way, high performance of the LED source (very stable emission spectrum and optical output power) is achievable with confidence, which allows high-sensitivity spectroscopic measurements of multi-species (aerosols and gases).
We report in this paper on the demonstration of an LED-based cavity-enhanced albedometer for simultaneous in situ measurement of aerosol scattering and extinction coefficients on the exact same sample volume.The performance of the instrument was evaluated using both laboratory-generated particles and ambient aerosols.The cavity-enhanced albedometer holds great promise for high-sensitivity and high-precision measurement of ambient aerosol scattering and extinction coefficients (hence SSA determination) and for absorbing trace gas concentration.The instrument's sensitivity and specificity demonstrated in the present work shows its potential for field observation on different platforms (ground observation networks, aircraft mapping, etc.), by benefiting from its capacity of distinguishing between aerosol extinction and trace gas absorption.In addition, simultaneous measurements of aerosol scattering and extinction coefficients enable a potential application for the retrieval of particle number size distribution and for faster retrieval of aerosols' complex RI.Moreover, unlike PAS technique, the measurement methods employed by the present albedometer are not (or much less) affected by RH, and hence well-suited to the measurements of aerosol optical properties at high RH, in particular for the determination of the complex RI of light-absorbing aerosols (such as black carbon and brown carbon) at high RH.
Currently, only one scattering coefficient can be measured due to the use of a single-channel PMT.When replacing this single-channel PMT with a multichannel PMT or a high-sensitivity spectrometer, measurement of broadbandwavelength-resolved scattering coefficients could be achievable.Employing a multi-cavity configuration could allow the albedometer to work in a wider wavelength range, from the UV to the near IR.

Fig. 3 .Figure 3 .
Fig. 3. Size dependence of the truncated fraction of total scattering under different truncation 14

Fig. 4 .
Fig. 4. Calibration of the scaling factor K' of the cavity enhanced albedometer for the 11

Figure 4 .
Figure 4. Calibration of the scaling factor K of the cavityenhanced albedometer for the scattering channel with He, N 2 , CO 2 and SF 6 at λ = 470 nm.(a) Plot of I scat /I tans ratio vs. theoretical value of the Rayleigh scattering coefficient of each gas.(b) Regression plot of the measured extinction and scattering coefficients for calibration.
Time series of a 5.5 h measurement of a particle free zero air sample with a time olution of 9 s (upper panel) and corresponding Allan deviation plots (middle panel) for th the scattering and extinction channels.The lower panel shows frequency distribution of performed scattering and extinction measurements.A normal distribution was fitted to the tograms.The 1 standard deviation, sd, is a measure of the instrument precision; and mean notes the mean scattering or extinction coefficients.

Figure 5 .
Figure5.Time series of a 5.5 h measurement of a particle-free zeroair sample with a time resolution of 9 s (upper panel) and corresponding Allan deviation plots (middle panel) for both the scattering and extinction channels.The lower panel shows frequency distributions of the performed scattering and extinction measurements.A normal distribution was fitted to the histograms.The 1σ standard deviation, SD, is a measure of the instrument precision; "mean" denotes the mean scattering or extinction coefficients.

Figure 6 .
Figure 6.Laboratory assessment of the particle loss vs. particle size in the albedometer developed.

Fig. 7 .Figure 7 .
Fig. 7. Time response of the developed cavity enhanced albedometer to a variation of part 17 Regression plot of the measured extinction and scattering coefficients, (b) ing and extinction coefficients as a function of particle concentration, and (c) the ing (Q Scat ) and extinction (Q Ext ) efficiencies as a function of particle diameter for isperse PSL spheres with four different particle diameters (200, 240, 300 and 400 nm) 470 nm.

Figure 8 .
Figure 8.(a) Regression plot of the measured extinction and scattering coefficients, (b) scattering and extinction coefficients as a function of particle concentration, and (c) the scattering (Q Scat ) and extinction (Q Ext ) efficiencies as a function of particle diameter for monodisperse PSL spheres with four different particle diameters (200, 240, 300 and 400 nm) at λ = 470 nm.

Figure 9 .
Figure 9. Survey of the measured values of real and imaginary part of refractive index versus wavelength for PSL.

Figure 13 .
Figure 13.Correlation plots between NO 2 mixing ratios measured with the cavity-enhanced albedometer and an NO x analyzer.All data were 5 min averaged.
Fig. 14.(a) Enlarged drawing of the scattering and extinction measurements of an air sampl 7