There exists a lack of aerosol absorption measurement techniques with low
uncertainties and without artefacts. We have developed the two-wavelength
Photothermal Aerosol Absorption Monitor (PTAAM-
Absorption of sunlight by aerosols, especially black carbon, and the related
(semi)indirect effects of these aerosols are collectively the second leading
cause of atmospheric warming (Bond et al., 2013). The same light-absorbing
aerosols have serious detrimental health effects (Janssen et al., 2011, 2012). They can be directly emitted into the atmosphere
(primary aerosols – black carbon and a fraction of the light-absorbing
organic aerosols) or produced in the atmosphere from precursor gases
(secondary aerosols – some of the light-absorbing organic aerosols) and
their optical properties changing with ageing (Saleh et al., 2013; Kumar et
al., 2018). Their short atmospheric lifetime (compared to CO
The absorption coefficient has been most often determined using the most practical instrumentation – filter absorption photometers: the Aethalometer, the particle soot absorption photometer (PSAP), the continuous light absorption photometer (CLAP) and the multiple angle absorption photometer (MAAP) (Rosen et al., 1978; Drinovec et al., 2015; Bond et al., 1999; Ogren et al., 2017; Petzold et al., 2002), where the measurement of light attenuation in the sample-laden filter relative to the clean filter is converted to mass equivalent black carbon concentration (BC) (Petzold et al., 2013a) or the absorption coefficient. This type of measurement requires different assumptions which are hard to verify. The enhancement of absorption in the scattering filter matrix and the mass absorption cross sections are two external parameters which need to be determined. Additionally, the measurement is non-linear, and the reduction of the sensitivity due to filter loading and the cross-sensitivity to scattering are artefacts, the contributions of which are either assumed (Bond et al., 1999; Ogren et al., 2017) or measured to different extents (Petzold et al., 2002; Drinovec et al., 2015). Post-processing algorithms were developed to compensate for the loading or concentration effects (Weingartner et al., 2003; Virkkula et al., 2007; Collaud Coen et al., 2010; Hyvärinen et al., 2013). These effects are, to a degree, sample dependent, so the data analysis requires more sophisticated approaches (Drinovec et al., 2015, 2017), and an additional measurement of scattering is needed (Arnott et al., 2005; Ogren et al., 2017; Yus-Díez et al., 2021) to fully compensate for the sensitivity dependence on the sample single-scattering albedo (SSA) and other artefacts.
In situ methods (i.e. without collecting the sample on a filter) offer advantages over filter photometers, such as separate measurements of aerosol extinction and scattering coefficients, where the absorption coefficient is calculated as the difference between these two parameters, hence the name of the method extinction minus scattering (EMS). The integration of these two measurements in a single instrument (Petzold et al., 2013b) allows the measurement of these quantities for the exact same sample, and calculation of their difference yields the absorption coefficient. Thus, this method was used as a reference for filter photometer characterization (for example, Bond et al., 1999). However, this measurement is still severely restricted when measuring aerosol exhibiting high SSA, as measurement errors of several percent can result in absorption coefficient errors exceeding 100 %, which requires adherence to a strict measurement and data post-processing algorithm to minimize these errors (Modini et al., 2021). On the other hand, the fractal nature of low-SSA black carbon aerosol and the effect of this morphology on the scattering truncation correction of the nephelometer present a systematic source of error also for measurements at low SSA (Modini et al., 2021). The large errors in the determined absorption coefficient at high SSA limit the use of EMS in regional and background sites, which are necessary for the quantification of the influence of absorbing aerosols on the climate.
A direct measurement of the aerosol absorption coefficient would therefore avoid the described issues. This is possible in photoacoustic instruments – photoacoustic spectrometers (PAS), in which an air sample is drawn through the sample chamber, where it is illuminated with a powerful pump laser beam. The light absorbed by the aerosol is converted to heat, causing an increase in local temperature. Transport of heat causes changes in density, and the acoustic wave is amplified in a chamber that acts as an acoustic resonator. The pump beam is modulated, and phase-sensitive detection is used to amplify the signal over the measurement noise. Acoustic resonance amplifies the signal by orders of magnitude.
Photothermal interferometry similarly employs a modulated pump laser to heat the sample with the absorbed laser light. A second interferometric laser probes the change of the refractive index caused by light absorption and the subsequent heating and decrease in the density of the sample. The response of photothermal interferometer (PTI) to the aerosol absorption coefficient is linear. The first proposed and realized PTI instruments employed a folded Jamin interferometer due to its inherent mechanical stability and a glancing pumping geometry, where the pump beam and the probe interferometric beam overlap at an acute angle in the sample (Moosmüller and Arnott, 1996; Sedlacek, 2006; Sedlacek and Lee, 2007; Lee and Moosmüller, 2020). A novel folded Mach–Zehnder interferometric design employs a single laser for both heating and interferometric detection (Visser et al., 2020).
A photoacoustic spectrometer is conceptually similar to the PTI; both use optical excitation to heat the sample. The difference is that a PAS employs an acoustic resonator, while PTI uses an optical resonator – the interferometer. PAS needs to actively track the acoustic resonant frequency to maintain maximum sensitivity, otherwise it can experience drastic changes in sensitivity. PTI maintains the maximal sensitivity by keeping the interferometer in the quadrature point, where the sensitivity is linear with phase and the change of phase is proportional to the normalized difference of the two intensities measured by the two detectors in the Mach–Zehnder (Visser et al., 2020; and our study) or Jamin (Sedlacek, 2006) interferometers.
The photoacoustic instruments may experience systematic biases when the
sample contains semi-volatile organic coatings or water. As these substances
evaporate from the aerosol, the latent heat causes a reduction of the
detected acoustic signal (Arnott et al., 2003; Murphy, 2009; Langridge et
al., 2013). As shown theoretically in Moosmüller et al. (2009), the
evaporation of semi-volatile species from the particle phase reduces the
sensitivity in PAS only for particles larger than a few micrometres using
typical photoacoustic design/operating parameters. We have shown in
experiments with coated soot that PTI and PAS instrumentation agree
(Kalbermatter et al., 2022) and that evaporation does not play a role in
investigation of soot particles coated with
Measurements of the effect of absorbing aerosols on the (regional) climate cannot be performed satisfactorily without reference methods. Reference instruments can be used to calibrate simpler but ubiquitous instrumentation. Reference instrumentation needs to measure the aerosol absorption coefficient directly, accurately and precisely. The vast majority of field instrumentation for the determination of the aerosol absorption coefficient is filter photometers. It is thus crucial to calibrate their response when converting the measured attenuation coefficient into absorption and to determine the limitations of filter-based measurements. This is especially so at (regional) background measurement sites, where climate change is monitored and where measurements are taken at high SSA – this is a limitation for filter photometers (Yus-Díez et al., 2021) and also for extinction minus scattering (Modini et al., 2021).
Standardization, validation and calibration of the aerosol absorption coefficient measurement require well-defined reference samples: particles or absorbing gases (Arnott et al., 2000). The greatest advantage of in situ methods is the ability of calibration with gases (Arnott et al., 2000; Lack et al., 2006, 2012; Nakayama et al., 2015; Davies et al., 2018), while filter photometers need to be compared with each other (Müller et al., 2011; Cuesta-Mosquera et al., 2021) or with other reference instruments (Bond et al., 1999; Arnott et al., 2003) using model aerosols: fullerene soot, spark discharge particles, soot generated with controlled gas combustion and dyes, especially nigrosin (Bond et al., 1999; Schnaiter et al., 2006; Müller et al., 2011; Cuesta-Mosquera et al., 2021). The advantage of using gases for calibration is conceptual – gas concentration can be measured with great precision and accuracy, and its absorption cross section is usually well known. It is also practical – gas calibration can be performed in the field, while instruments calibrated with aerosolized particles need to be sent to a calibration laboratory as this effort requires a sophisticated setup.
Calibration with NO
This work describes a novel photothermal interferometer design, which allows for simultaneous and colocated absorption measurements at two wavelengths. The article has been divided into a number of sections to provide a thorough treatment of the instrument: firstly the physical construction of the instrument is described, with a special focus on the novel geometry of the pump and probe beams in the sample chamber. The temporal evolution of the signal measured by the photodiodes during the modulation period of the pump beam and the selection of the modulation frequency are then discussed. Subsequently a noise analysis is performed, showing the sources of noise in the instrument. Following this the operation of the interferometer and the interferometric signal is detailed. A section of the article is then dedicated to the two-wavelength calibration procedure. The linearity of the measurement is demonstrated, and the stability and noise of the instrument are determined. Finally, the results obtained with the instrument during a number of laboratory and ambient campaigns are presented.
Nigrosin (Acid black 2, Nigrosin water soluble, CAS 8005-03-6) and ammonium sulfate (Mascagnite, ReagentPlus®, CAS 7783-20-2) were obtained from Sigma-Aldrich. Aqueous solutions of nigrosin (labelled, N1–N4) and ammonium sulfate (AS1–AS3) were prepared by dissolving chemicals in ultrapure water (Milli-Q) (Table 1).
Designations of the aqueous solutions of nigrosin (N1–N4) and ammonium sulfate which were nebulized.
Solid nigrosin samples were prepared on microscope slides by drying the
aqueous solutions of nigrosin (see sample photographs in Fig. S5.3 in the Supplement).
Glass slides were first cleaned with isopropyl alcohol, and then a hydrophilic
layer was produced by dipping the slide into a 4 % detergent solution
(Hellmanex III) for 30 s and flushing the slide with Milli-Q water.
Finally 0.2 mL of nigrosin solution was spread on a surface of approx. 6 cm
Two bottles of 1
Poor stability of reference gas mixtures in the nmol mol
First, the permeation rate of the permeation unit (produced by Fine
Metrology S.r.l.s., Italy) containing high-purity NO
Nigrosin and ammonium sulfate aerosols were generated using the ATM 226
nebulizer (Topas GmbH, Germany) set to 4 L min
Diesel exhaust was collected from the tailpipe of a EURO3 Volkswagen Passat
1.9 TDI at 2500 rpm. An average black carbon emission factor of 1.31 g kg
Soot generated by a miniCAST 5201 Type BC (Jing Ltd., Switzerland) was also used as a test aerosol. The physical and optical properties of the soot particles have been reported elsewhere (Ess et al., 2021).
Propane soot samples were obtained by collecting the exhaust of a portable propane burner/torch. After ignition the air inlets were closed to generate fuel-rich combustion. Because of slow sample dilution large soot agglomerates were formed.
Absorbance of nigrosin was measured using a Shimadzu UV-3600 UV-VIS-NIR spectrophotometer (Shimadzu, Japan). Aqueous solutions of nigrosin were measured in the 10 mm path quartz cuvette. Solid nigrosin samples were placed at the entrance of the ISR-3100 integrating sphere, which collected all the transmitted illumination.
Reflectance of nigrosin using a perpendicular beam at 520 and 1064 nm was measured to correct for the reflection losses during the absorbance measurements. Reflectance was calculated by dividing the intensity of the reflected beam with that of the incident beam.
To determine the real part of the refractive index, a measurement of the
Brewster
The size distribution of the aerosol particles was measured using a scanning mobility particle sizer (TSI model 3936L75). The instrument was used with an impactor with a 0.0508 cm nozzle. Measured spectra were corrected using the diffusion and the multiple charge correction algorithms. Scattering was measured with the Aurora 4000 polar nephelometer (Ecotech, Australia) set to measure the total and back-scattering coefficient.
Two filter photometers, an Aethalometer model AE33 (Magee Scientific, USA)
and a CLAP (Haze Instruments, Slovenia), were used to obtain attenuation
coefficients. The AE33 features a built-in filter loading compensation
algorithm (Drinovec et al., 2015) and was using the M8060 filter. The CLAP
was run using the Azumi 371M filter, and the data have been compensated for
manually using the algorithm by Ogren et al. (2017). For comparison with
PTAAM-2
Two NO
To determine the morphology of the diesel exhaust, propane soot and nigrosin samples, a FEI HeliosNanolab 650 (Thermo Fisher Scientific, Waltham, MA, USA) scanning electron microscope (SEM) operating at 1 kV accelerating voltage was used. The samples were mounted on SEM holders with carbon tape and coated with a few nanometres of carbon to prevent charging effects.
To determine the laser stability during warmup and operation, the spectra of the 532 nm DPSS laser were measured with a spectrometer (Shamrock SR-500i, Andor) equipped with a cooled EMCCD camera (Newton DU970N, Andor).
An atomic force microscope (AFM) Nanoscope IIIa – MultiMode AFM (Digital
Instruments, Santa Barbara, CA) equipped with J scanner (100
Mie calculations were performed using MATLAB routines for homogeneous spheres (Mätzler, 2002).
The presented data were collected during several measurement campaigns:
The experiment setup for Ljubljana campaigns is presented in Sect. S7 in the Supplement.
The dual-wavelength photothermal aerosol absorption monitor (PTAAM-2
Schematic representation of the PTAAM-2
The sample cell contains pressure, temperature and relative humidity sensors. The interferometer is sealed in the enclosure to protect it from acoustic noise and pressure fluctuations. The optical path length of one of the interferometer beams is controlled by a pressure cell connected to a computer-controlled syringe pump.
A 2 W frequency-doubled Nd:YAG laser (532 nm) and a 3 W Nd:YAG (1064 nm) laser are coupled to the multimode optical fibres. Parts of these fibres are shaken by an electromagnetic actuator to homogenize the profiles of the beams exiting the fibres. The pump beams, which are modulated at different frequencies, are collimated and introduced by dichroic mirrors into the interferometer. The powers of the pump beams are monitored with photodiodes P1 and P2.
Our photothermal interferometer configuration utilizes an axicon to focus the pump beams into the sample chamber where they overlap with the probe beam (Drinovec and Močnik, 2020). The pump beams exiting the fibres are collimated and subsequently combined using two dichroic mirrors and then directed into the measuring chamber using a custom drilled mirror – two drilled holes allow the interferometer beams to pass through the mirror without interacting with the pump beam. A custom axicon, drilled to allow passage of the probe beams, is used to provide an elongated focus of the pump beams along the axis of the probe beam inside the sample cell. The diameters of the probe and pump beams in the sample cell are shown in Fig. 2. The pump beams are aligned with the probe beam using a CCD camera. During 1.5 years of instrument testing (including road shipment in excess of 3000 km to the two measurement campaigns), there was no need to realign the optics.
Beam diameters of probe and pump beams inside the sample cell. Beam sizes closer to the axicon could not be measured due to mechanical restrictions.
This setup allows for simultaneous measurement of absorption at two wavelengths in the same sample volume by modulating the probe beams at different frequencies (91 and 96 Hz, for example). The retention time of the aerosol particles in the pump beam (approx. 5 s for a 100 nm particle) is much longer compared to the duration of the modulation interval (11 ms).
For “in situ” aerosol absorption measurement the pump beam intensity
should be kept low enough so that the heating does not modify the physical
properties of the particles. The use of an axicon for focusing the pump
beams results in a pump illumination intensity of approx. 2 W mm
When the sample air is illuminated with the pump beam, some of the light is absorbed by the gases and particles in the sample. Heat is then transferred to the surrounding air, causing an increase in temperature. This, in turn, causes a small reduction in the air refractive index. This change of the refractive index is measured as the reduction of the optical path in the folded Mach–Zehnder interferometer.
The pump beam is modulated ON or OFF with a 50 % duty cycle. When the pump beam is switched ON, there is first a linear increase in the photodiode voltage, which later starts to saturate due to increasing heat transfer to the surrounding air (Fig. 3). The exact shape of the detected voltage depends on the intensity profile of the pump beam during the heating period and the geometry of the pump and probe beams. For the pump frequencies between 90 and 100 Hz, we observed an exponential rise and decay of the detector voltage during the ON and OFF part of the modulation period. Here, we show the results obtained for nigrosin using pump lasers operating at 532 and 1064 nm. The amplitude of the 532 nm channel voltage is approximately 7-fold higher compared to the 1064 nm channel, but the shape does not change with amplitude. Therefore, nonlinearities in the instrument response can be avoided, and a lock-in amplifier can be used to measure the amplitude of the photodiode voltage.
Average photodiode voltage during the heating and cooling parts of the modulation period for the 532 and the 1064 nm channels obtained for nigrosin particles. The duration of the modulation period is 11.1 and 10.4 ms, respectively. The curves are obtained by averaging the photodiode PD1 voltage over more than 10 000 modulation periods.
When the optical path of the interferometer changes, we observe a voltage
increase on one of the photodiodes and a decrease on the other. To increase
the signal-to-noise ratio, the difference between the photodiode voltages
(
Instrument response
Increased noise at lower frequencies can also be observed on fast Fourier transformation of the photodiode signal (Fig. S3). Noise is highest below 10 Hz; there is a plateau between 10 and 100 Hz and increased noise between 200–300 Hz. There are also certain noise peaks, which should be avoided to reduce the measurement noise. Frequencies of 91 and 96 Hz were selected to attain the highest signal-to-noise ratio.
When there is no absorbing sample in the measurement cell, it is possible to
measure an offset with the magnitude of 8
There are several sources of noise which are independent of the pump
frequency, including the noise of probe lasers, photodiodes, sample air flow
turbulence, pump laser power oscillations, mechanical resonances of the
optical elements and random electronic noise. The sources of noise were
investigated by comparing different experimental setups (Fig. 5).
The results show that the majority of the noise comes from the
interferometer. It may be related to the vibration of the optical elements
of the interferometer. For the 1064 nm channel about one-half of the noise
is caused by the offset variation.
Contribution of different sources to the measurement noise
(standard deviation of signal) for the 532 nm channel
The response of the photothermal interferometer is highest when the
interferometer is operating in the so-called quadrature point – where the
difference of the optical paths between the two arms of the interferometer
is exactly
The interferometer scan of photodiode PD1 and PD2 voltages is
performed by changing the optical path of one of the interferometer arms
using the pressure cell. A
For an interferometer in the quadrature point, the lock-in signal is
proportional to the amplitude of the interferometer signal
The instrumental response depends on the overlap between probe and pump beams. After beam alignment, both channels need to be calibrated.
Similarly to the calibration of photoacoustic instruments (Arnott et al.,
2000; Lack et al., 2006; Bluvshtein et al., 2017), we decided to use
NO
Calibration of the 532 nm channel was performed by filling the Tedlar bag
(Sigma-Aldrich) with NO
The measured NO
Calibration of the 1064 nm channel is more complicated as there are no
appropriate absorbing gases available at that wavelength. Particles with
known optical properties can be used here to transfer the calibration from
the NO sphericity of nigrosin particles confirmed by scanning electron microscopy
(Sect. S6), nigrosin particle size distribution measured with SMPS, the complex refractive index of solid nigrosin at 532 nm and 1064 nm.
With respect to the nigrosin refractive index investigation, the imaginary
part can be determined from its absorbance. Absorbance measurements of both
aqueous nigrosin solution and nigrosin film were conducted (Sect. S5.2
and S5.3). The absorbance of the aqueous solution of nigrosin was measured
in a 1 cm path cuvette. Solid nigrosin samples were produced by drying
nigrosin solution on the microscope slides. The absorbance of the solid
nigrosin film on these slides was measured with an integrating sphere
spectrometer. The results show large (up to
The imaginary part of the refractive index
The real part of the refractive index was determined by the Brewster angle
measurement on nigrosin film (Sect. S5.1). Measured values are on
average 0.03 lower compared to Bluvstein et al. (2017) (Fig. 7b), which is
just outside the measurement uncertainty of 0.02. The following values of
the refractive index have been used for Mie calculations:
The absorption ratio
During different measurement campaigns the absorption ratio was calculated
for the measured nigrosin size distributions. Due to the increased
measurement uncertainties for particles bigger than 400 nm, these data were
not used for Mie calculation. Absorption ratio obtained with sample N2
ranged between 0.0781 and 0.0785 (Sect. S5.5). The calibration
parameter
The 532 nm channel of the PTAAM-2
Mie calculated and measured absorption and scattering coefficients at 532 nm for aerosolized nigrosin N2 for several independent experiments (designated A, B and C) during the AeroTox 2020 and Ljubljana 2021 campaigns. Size distribution below 400 nm was used for Mie calculation. Experiment mean values and standard errors are presented.
Before conducting the measurement campaigns, the PTAAM-2
Validation of the PTAAM-2
Absorption measurements with the PTAAM-2
Average value and standard deviation of the offset and signal for
filtered air
The linearity of the instrument response was tested with absorbing gases and particles. For NO
Instrument response to different amount fractions of NO
Interestingly, the noise behaviour for gas measurements differs
substantially from that of particle measurements (Fig. 12). There is no
significant increase in noise (standard deviation at 1 s time resolution)
with the NO
Standard deviation of the 532 nm channel data during measurement
of NO
Because scattering does not produce thermal effects, we do not expect
photothermal instruments to be sensitive to the scattering of the sample. We
generated purely scattering aerosol by nebulizing different aqueous
solutions of ammonium sulfate (Fig. 13). For the 532 nm channel, we observed
a constant signal of about 1.5 Mm
Scattering artefact for different samples of ammonium sulfate
As shown, the PTAAM-2
Scattering artefact of filter photometers for CLAP and the AE33
for different sizes of ammonium sulfate particles
Similarly to scattering, there is also a strong dependence of the filter
photometer response on the size of absorbing particles. The enhancement of
light absorption in the filter matrix relative to particles suspended in air
is described by the multiple-scattering parameter
The dependence of the parameter
The AE33 systematically overestimates the absorption
Ångström exponents (AAE) compared to the PTAAM-2
The dependence of the AE33 measured attenuation coefficient on the sample
single-scattering albedo (SSA) was tested with an external mixture of
propane soot (441 nm volume mode) and ammonium sulfate (123 nm volume
mode). The experiment was performed by filling a barrel with soot at the
start, followed by continuous injection of ammonium sulfate particles (Fig. 15). SSA increased from 0.25 to 0.997. There is an increase in attenuation ratio
Measurement of externally mixed diesel soot and ammonium
sulfate. The ratio between the AE33 attenuation coefficient and the
PTAAM-2
The winter campaign was carried out in Ljubljana (Slovenia) in February and
March 2020 to evaluate the PTAAM-
Absorption coefficient measured during the Ljubljana 2020 winter
campaign
The uncertainty of the measured absorption coefficient results from the
calibration, method and instrumental uncertainties (Table 4). The
calibration of the 532 nm channel depends strongly on the uncertainty of
NO
The calibration of the 1064 nm channel depends on both the uncertainty of
the 532 nm channel and the uncertainty of the calculated nigrosin absorption
ratio
Combined standard uncertainties for the determination of absorption coefficients and absorption Ångström exponent are presented in the lower part of Table 4. The 1064 nm channel uncertainty (6 %) is higher compared to the 532 nm channel (4 %) due to the additional calibration step with nigrosin particles. The uncertainty of the absorption Ångström exponent (9 %) is higher compared to the absorption coefficients because of the properties of the logarithmic function. Ångström exponent depends mostly on the correct determination of nigrosin refractive index at the measurement wavelengths.
The sources of uncertainty for PTAAM-
We report on the design and operation of the first dual-wavelength
photothermal interferometer (PTAAM-
The instrument was calibrated with NO
The PTAAM-2
We believe that the demonstrated operation and performance make the
PTAAM-
The raw data and measurement logs are available at
The supplement related to this article is available online at:
LD and GM designed and developed the PTAAM-
Luka Drinovec, Griša Močnik and Uroš Jagodič are or were employed by Haze Instruments d.o.o., the manufacturer of the described instrument. Technologies described here-in have been protected with patents.
Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank Martin Gysel (PSI) for the use of the AE33, Teledyne API and EAS
Envimet Analytical Systems GmbH for the loan of the NO
This research has been supported by the Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (grant no. 200021_172649), Eurostars (grant no. 11386), the Javna Agencija za Raziskovalno Dejavnost RS (grant nos. P1-0385, P1-0099 and I-0033), and the European Metrology Programme for Innovation and Research (EMPIR Black Carbon and EMPIR AeroTox grants).
This paper was edited by Mingjin Tang and reviewed by three anonymous referees.