The sheath and input flows of the NAIS are critical for a precise determination of the particle mobility and concentration. Thus, prior to the mobility and concentration calibrations the instrument should be cleaned, leak tested and flow checked. The cleaning procedures are essential before the determination of flows. If dirt enters the tubing or nets inside the instrument, the flow resistance will alter the volume flow through the Venturi tubes (see Sect. 3.5.2). When maintenance cleaning is done regularly, the instruments can perform well for extended periods and flows stay stable (Gagné et al., 2011). This is primarily relevant for the instruments with the 1-blower flow system as the correct flow balance is very delicate (see Option A). The newer instruments with three and four blowers will maintain flow stability for much longer periods without the maintenance and will only require recalibration in case a simple check indicates a problem (see Option B).

Leak tests can be done using several methods. Alternative 1: the volume flow rate measured from the inlet and outlet should match when no leaks exist. This is sufficient in case the instrument is operated at atmospheric pressure and a low pressure drop inlet is used. Alternative 2: this alternative is carried out by blocking the inlet and outlet, applying overpressure or under pressure inside the instrument, and measuring the flow rate required to maintain the constant pressure. The recommended range is 50–100 mbar overpressure or under pressure. Overpressure above 200 mbar may damage the instrument. Alternative 3: when the inlet is closed, the recorded sample flow rate should drop to a value well below 10 L min-1. This method is suitable for a quick check; however it is not completely reliable.

The flow calibration should be done once a year during the long-term operation, or before and after a short-term campaign measurement. However, the flows should be determined always, when a blower is replaced or a large leak is detected and sealed.

For the 1-blower system, in Fig. 1a, one blower runs all the flows. Five flow rates (sheath and output flows of both analysers and the total exhaust flow) are measured with Venturi tubes. In normal operation only the exhaust flow rate is measured continuously, as it is the most sensitive to the changes of all the other flows. Each Venturi tube has an individual calibration where an exact pressure drop, corresponding to the specified flow rate, is determined in normal conditions. The Venturi calibration values are provided by Airel Ltd, when the instrument is manufactured. The Venturi calibration is not needed to be done by the user, but the pressure drop should be verified and the flow adjusted, if needed. This procedure is illustrated in Fig. 5. The differential pressure over the five Venturi tubes should be measured with a reference differential pressure measurement device (e.g. TESTO 512 0–2 hPa), and checked against the value obtained from the calibration to be sure that the flow rate though the Venturi is correct. If the checked pressure difference does not correspond to the calibrated ones, the flows should be adjusted to match the calibrated pressure difference while keeping a constant overpressure (80 mm H2O) after the blower (measured with, e.g. TESTO 512 0–20 hPa).

The procedure for determining the flows for the 1-blower system is as follows: first, the user verifies that the sample and outlet flows are equal and no leaks exist. Second, the blower power should be adjusted so that the overpressure after the blower is 80 mm H2O. Third, each of the five flows are measured via the pressure difference from the Venturi tubes, and adjusted, if required, to match the values given in the Venturi calibration (while keeping a constant overpressure after the blower). To adjust the Venturi's pressure drop turn the brass knob which is located on a side of the flow adjusting valve with pliers (middle bottom panel, Fig. 5). To fix misbalanced flows, either the valve on the higher pressure drop side should be slightly closed or the other valve slightly opened. Last, the user should check that the sampling flow remains at ∼ 60 L min-1. For a full pressure sensor calibration the user should contact Airel Ltd. to provide detailed information on the flow calibration set-up and possibly update the measurement software.

For the 3- and 4-blower system, in Fig. 1b–c, the verification involves measuring the volumetric flow rates with an external flowmeter (e.g. TSI 4000 series) at the instrument's exhausts and after the sheath air filters. First, place a reference flowmeter at both exhausts (measure one exhaust flow tube at time) to verify the sample flow at the Venturi tubes, which are located just before the exhaust tubes. The sample flow and exhaust flow should be identical when the instrument is not leaking. Second, disconnect a sheath flow tube and place the flowmeter after the sheath air filter to verify the sheath flow at Venturi tubes, which are located prior to the blower and filter. Figure 6 shows a quick and easy way to check the sheath flow of the negative polarity for the 4-blower system. Repeat this step for the sheath flows for the both polarities. For a simple instrument check it is sufficient to compare the reference flowmeter value to the corresponding NAIS flowmeter value. Note that the NAIS flowmeters show the actual volume flow rate that is not adjusted to standard temperature and pressure conditions. A full flow sensor calibration can be done by the user but it requires assistance from Airel Ltd. to provide a customized calibration software, followed by processing the results and updating the measurement software.

In case of the 3- and 4-blower systems, the instrument includes a barometric pressure sensor that is actively used to determine the correct flow rate. The instrument sensor value should be compared to a reference barometric pressure sensor and the calibration coefficients should be adjusted, if necessary. To change these calibration coefficients, contact to Airel Ltd.

In the case of the 3-blower system, there is only one sample flow Venturi sensor in the instrument that measures the total flow from both analysers. An additional step is required to confirm the sample flow balance of the negative and positive analysers. This involves adjusting the two valves before the Y-connector, where the sample flows join from both of the analysers. The two Venturi tubes next to the valves should be measured simultaneously using two handheld differential pressure sensors. The calibrated values for these pressure differences are either written inside the instrument or available from Airel Ltd. The valves should be adjusted accordingly.

The response of the DMA high voltage (HV) supply should be followed from the instrument diagnostics. Correct sizing of small ions and particles in the DMA is highly sensitive to the accuracy of the applied HV. Particular care is required particularly in the low voltage range, which is used to classify the smallest ions. The voltages in the inner electrode are ±9, ±25, ±220, and ±800 V, depending on the polarity of the DMA. For the newer versions of the NAIS manufactured after 2014 (serial numbers NAIS27 and NAIS-4-1, and later) the corresponding voltages are ±9, ±35, ±150, ±700 V. To confirm the voltages with an independent measurement, a HV-probe with ultra-low impedance should be used.

In the size range of the cluster ions and small neutral particles the calibration is a challenging task due to limitations in the capability of reference instruments, generation of proper calibration aerosols, and instrumentation for the size-separation. Possible calibration set-ups are presented in detail in Asmi et al. (2009), Gagné et al. (2011), Kangasluoma et al. (2013) and Wagner et al. (2016). To determine the losses and the sizing accuracy of the full size range of the NAIS requires an extensive suite of instrumentation: (1) in the sub-10 nm size range with a high resolution H-DMA (Herrmann-DMA; Herrmann et al., 2000; Kangasluoma et al., 2016) to determine the transfer functions and losses, (2) monomobile molecular standards (Ude and Fernández de la Mora, 2005) to determine specific mobility calibration, and (3) a Hauke DMA (Winklmayr et al., 1991) to perform the mobility and concentration calibrations in the size range from 4 to 40 nm. Prior to the loss and mobility calibration, the flows need to be verified as described above.

Wagner et al. (2016) studied the accuracy of the NAIS in a supplementary laboratory calibration. They concluded that in ion mode the sizing of the NAIS was very accurate, regardless of the version of the data inversion, and the ion number concentrations were underestimated 15–30 %, depending on the version of the data inversion. Using a correction introduced by Wagner et al. (2016), the uncertainty of the ion concentration measurement of the NAIS can be reduced to ∼ 10 %, allowing the NAIS to be used in quantitative ion cluster and charged particle studies.

It is recommended that different methods are used in parallel to determine the cluster and nucleation mode number concentration in order to avoid misinterpretation of results (Kulmala et al., 2012). Participation in the intercomparison workshop is recommended. Another opinion is to perform a side-by-side intercomparison at the measurement site, if a suitable supporting instrumentation is available (e.g. ion spectrometer, Gerdien counter, condensation particle counter, differential/scanning mobility particle sizer, or air conductivity measurement, e.g. Asmi et al., 2009; Gagné et al., 2011).

To verify instrument operation prior a measurement campaign, the balance of number concentration measured with the positive and negative columns should be checked. The concentration verification can be done by generating population of sample aerosol (in equilibrium charge distribution) and measuring it with both columns. A good agreement (10–20 %) between the polarities gives confidence that the instrument is working properly. In the following sections we describe three options for the verification experiments before instrument deployment.

Peeling a citrus fruit and thus releasing D-limonene, a common monoterpene, into the room air can lead to aerosol particle formation in the indoor environment (Vartiainen et al., 2006). This is a fast, easy and cheap way to generate nucleation mode aerosol particles over a whole size range of the NAIS as the D-limonene oxidation products trigger the particle formation and subsequent growth (Gagné et al., 2011). Finding the right amount of fruits to peel to generate the correct amount of vapour can take a few trials. Note that there should be sufficient ozone concentration and low background aerosol concentration in the room to facilitate new particle formation and growth.

Fast concentration verification between the polarities can be done by sampling from indoor and outdoor. The indoor sample due to efficient air-conditioning is typically dominated by cluster ions and can be used to check the balance between small ions, whereas outdoor sample has a typically abundant Aitken mode and can be used to test larger ions and particles (Hirsikko et al., 2007).

Extensive number concentrations in the range from ∼ 102–106 cm-3 in both positive and negative cluster ions can be generated by using an external radioactive source (Steiner et al., 2014). A radioactive source producing bipolar charger ions should be placed right in front of the NAIS inlet. By varying the flow rate through the external charger, the user can vary the number concentration of the charger-generated ions. This data verifies the balance between the positive and negative cluster ions and their proper size classification. Please note that typically the negative cluster ions have higher mobility than the positive cluster ions. The charger ion size distribution also varies as a function of carrier gas composition (Manninen et al., 2011).

A recommended inlet is a 50 cm-long metallic tube (diameter 35 mm) with a bend (90∘ angle facing down) and metallic net (grid size 1 mm) in the end of the inlet line; see Fig. 7. The brass inlet connector with a metallic net inside (sold by Airel Ltd.) is recommended to be used. Although the instrument and the inlet line can be placed vertically or horizontally, the horizontal orientation for the inlet line is recommended. In the vertical inlet set-up the precipitation may easily enter the instrument and damage the instrument and lead to poor data quality. Sampling height depends on the surroundings. It can vary from 2 to 15 m above ground level (a.g.l.; height of the surrounding canopy/buildings). Note that the Earth's electrode effect can cause an imbalance between polarities (negative and positive ion number concentrations), when the inlet height is below 4 m a.g.l. The ionosphere has positive charge and Earth's surface has a negative charge. Thus, Earth's surface works as a negative electrode which attracts the positive ions and repels the negative ions close at ground level (Hoppel, 1967).

The inlet lines should include a proper rain cover as the rain droplets can interfere the measured spectra (Tammet et al., 2009). In the field conditions, the user should make sure that rain does not get into the instrument enabled by high sampling flow rate. If there is a chance of water dripping into the inlet, the end of the inlet tube should point at least slightly downwards. A metallic grid in the inlet is recommended. The performance of the mobility analyser is sensitive to insects and other material which may settle on the electrodes or electrometers. These impurities can cause corona discharge, noise and parasitic currents. Furthermore, the pressure drop from the inlet to the instruments should be kept in the range of few hPa. This is facilitated by a regular cleaning of the inlet grid.

Due to diffusional losses of small particles the inlet lines needs to be kept as short as possible and as straight as possible, while keeping the flows close to laminar. Enhanced diffusional particle losses may occur in the sampling lines with bends or elbows. The particle losses increase with a decreasing radius of the bend. It is very essential that the inlet lines and connectors should be made from a conductive material (preferably stainless steel) to avoid losses caused by a static electric charge. Experience has shown that non-conductive tubing (e.g. plastics) may remove a considerable fraction of any charged particles by the unwanted electrostatic forces. A rough estimation for particle losses should be done on the measurement site after the installation by measuring with and without inlet set-up by performing a short measurement exercise with and without the inlet construction.

When working in a warm and humid atmospheric environment, dew point temperature of the sample flow can reached in the measurement cabin or container (20–25 ∘C). This requires that the aerosol sample flow has to be dried, either directly in the sampling line or at the instrument. During earlier deployments in South America (Backman et al., 2012) drying by heating has been used for the NAIS. The inlet line and at the instrument is heated above ambient temperature to avoid water condensing inside the tubing or the instrument. At high altitude sites and other sites with heavy snow storms heated inlet has been used to avoid ice blocking the inlet line. Heating of the sample flow may change the sample by evaporating the most volatile particulate species. To limit relative humidity in the aerosol sample flow, we do not recommend a membrane dryer (e.g. Nafion^™ dryer), or a silica-based aerosol diffusion dryer due to high aerosol sample flow rates and increased diffusional losses of clusters and small particles. When sampling in a highly polluted environment, we recommend adding a core sampling inlet and a dilution of the sample with aerosol-free bypass flow (drying by dilution; see Wagner et al., 2016).

In a typical field operation, the NAIS should be placed indoors in an air-conditioned space. The instrument should be operated in a temperature of 5–35 ∘C to avoid a malfunction of the electric circuits (chargers and filters, HV supplies). The instrument inlet and the immediate vicinity should be grounded well so that the sampled ions are not attracted by the static charge on nearby surfaces. The instrument itself should be grounded via the power cable. The measurement cycles should be set based on scientific aims, briefly summarized in the sections below.

In close to sea level, the NAIS software uses a fixed sheath-flow rate value of 60 L min-1 and a sample-flow rate value of 30 L min-1 (1-blower system) and 27 L min-1 (3- and 4-blower system) per DMA, when calculating the ion and particle number size distributions. The deviation of the flow rate from the default value should be taken into account during subsequent data processing by applying a correction to the number size distributions. The volumetric flow rates are typically recorded by the instrument. However the measured distributions are not automatically corrected in the case if the flow rates deviate from the nominal values. When the sample-flow rate is recorded, the ion and particle number concentrations can simply be multiplied by (default flow rate) / (recorded or measured flow rate) ratio. When the sheath flow has changed from default values the ion and particle size distributions should be re-inverted (assistance from Airel Ltd. is needed). For the ground-based measurements, the recommended measurement cycle is as follows: offset – 30; ions – 90; particles – 90.

At high-altitude sites, the NAIS (with 3- and 4-blower systems) volumetric sample flow rate is kept constant whereas the sheath-flow rate is varied automatically (Mirme et al., 2010). The automatic adjustment of the sheath-flow compensates changes in the particle mobility in exceptionally low or varying air pressures and temperatures. Thus, the classified particle size range is kept invariant of the pressure and temperature changes. The effect of ambient temperature variations to measured ion and charged particle mobilities is considered small because of warming in the sampling line. The 1-blower system is not recommended for high-altitude or aircraft measurements and we recommend upgrading it into the 3- and 4-blower system. Otherwise, the user needs to keep the blower operating in the right volumetric flow range manually or apply a correction to the number size distributions, and a separate airflow calibration is needed for the 1-blower system in the variable environmental conditions.

Several improvements were made to the airborne NAIS to able to measure the size distribution and concentration of ions as a function of altitude inside a pressurized aircraft (see Mirme et al., 2010). When measuring inside a pressurized aircraft, the instrument leaks have to be particularly well controlled. We recommend using the upgraded 3- and 4-blower system in the aircraft deployments. These versions of the NAISs automatically adjust the aerosol sample- and sheath-flow rates so that the particle sizing and volume sample-flow rate remain constant regardless of ambient pressure. Typical modifications to the NAIS in the airborne measurements are (1) replacing the electricity supply to match the system used in the aircraft, (2) designing a special inlet system to sample air from outside the aircraft, (3) reinforcing the instrument rack and attaching it into the aircraft frame, (4) modifying the instrument for increased air tightness in the case it is used in a pressurized cabin, and (5) setting the length of the measurement cycle to a minimum. If the inlet is in over pressure, the exhaust flow may need to be restricted with valves (e.g. adding some soft tubing and a pinch cock). For the airborne measurements the recommended measurement cycle is as follows: offset – 30; ions – 45; particles – 45.

In chamber measurements, where it is required to minimize the amount of sample air, the high sample flow rate of the NAIS is a challenge. In such an application, the it is recommended to operate the NAIS with a recirculation system, which dilutes the inlet sample flow with filtered air coming from the exhaust of the instrument. In other words, the sample air from the chamber is diluted with a portion of the exhaust air of the instrument, which is filtered with a high-efficiency particulate air (HEPA, e.g. Camfil Megalam, MD 14-305X305X66-10) filter and mixed with the sample air (in more details; see Franchin et al., 2015). The pressure drop on the filter and dilution mixer should be below 10 hPa to ensure that the sample flow blowers of the NAIS are able to comfortably circulate the air. The use of the dilution system allows reduction of the fresh sample flow from 54 to 20–30 L min-1.

Otherwise in the laboratory measurements, where the available sample flow rate is limited and sampled concentrations are small, it is recommended to use only one polarity (column) of the NAIS (e.g. Manninen, 2011) at a time depending on the polarity of the user needs. This is possible only with the 4-blower system, where the columns work completely independently. To disable one column, disconnect and close the corresponding tube at the inlet Y-connector which divides the flows to the two analysers or alternatively close the corresponding exhaust outlet. The instrument configuration files should be modified to switch off the blowers for the disabled column. This must be done to avoid damage to the blowers. When the NAIS is used in aerosol exposure studies with extremely high concentrations (∼105–106 cm-3), we recommend adding a core sampling inlet with appropriate dilution as well.

The instrument operation should be checked daily by the user. Figure 8 summarizes a recommended checklist for instrument's performance monitoring. Airel Ltd. provides tools for this. There are two programs in the measurement software package provided with the NAIS: (1) Spectops for running the measurements and viewing online data and (2) Retrospect for viewing and reprocessing the recorded data offline. The Airel Ltd. has an extensive diagnostic checklist in the NAIS manual: http://wiki.airel.ee/Docs/NaisManual. They are not repeated here, but we recommend the user to follow the checklist.

The blowers are automatically adjusted using a software digital feedback loop controller based on the flow sensors and barometric sensor so that the mobility analysis and sampling volumetric flow rate remain constant. When measuring through an inlet with a high pressure drop or in polluted conditions, the blower might be under heavy strain as this requires the blower to be operated with a higher voltage. A similar situation is reached, when the flow resistance increases over time due to settled atmospheric aerosols (instrument getting dirty). This calls upon a frequent cleaning of the instrument in the polluted conditions.

To keep the corona charger efficiency at a constant level independent of environmental conditions, the corona needle voltage is adjusted by varying the HV supply feed voltage according to an active feedback loop. The efficiency of the corona charger is directly determined by the charger ion concentration, which is proportional to the electric current carried by the corona ions to the surrounding electrode inside the charger volume. Thus, the discharger and the main charger are controlled with a feedback loop driven by the current measured from the surrounding electrode. In normal operating conditions of the NAIS, the corona-needle voltage is in the range of 2–3 kV. Over time the electrode (i.e. corona needle/wire) gets worn out because of aerosol particles settling on the needle. This will typically cause the corona voltage fluctuate as the corona ignition voltage increases and a stable discharge can no longer be maintained. For this reason, the NAIS corona needles need to be cleaned regularly.

In the particle mode the post-filters remove the cluster ions generated by the corona chargers. The post-filter settings should be optimized according to environmental conditions. Typically the filter operates with a voltage of 40–100 V. The corona-generated ions are at the same size range as the smallest particles measured by the NAIS. The set point of the post-filtering voltage is a compromise between the removal of corona-generated ions and the penetration of small aerosol particles. The main electrical filter voltages are adjusted by varying the HV supply feed voltage according to an active feedback loop. In the 3- and 4-blower systems, the user can change the target values by modifying the configurations with Spectops software (Mirme and Mirme, 2013). In the 1-blower systems, the post-filter voltage is adjusted by the user manually; see Supplement S2 for details. Figure 9 shows when to change settings or adjust manually the post-filter during the particle measurements. We recommend that in the continuous field measurements the automatic adjustment is used, if possible. In the laboratory experiments with rapidly changing or unusual aerosol distributions, the automatic adjustment should be switched off (Manninen et al., 2011).

The inlet net and inlet tubing should be cleaned thoroughly in 1–3 week intervals to maintain the optimal aerosol sample flow and reduce the amount of dirt settling on the analyzer. The required interval depends on the local aerosol concentrations.

During long-term operation the NAIS should be cleaned every 1–3 months due to the deposition of particulate matter inside the instruments. Within the cleaning procedures all parts, which are in contact with the sample- and sheath-flow, should be thoroughly wiped using delicate task disposable wipes (e.g. Kimberly-Clark Kimtech Science Kimwipes) and alcohol (e.g. 2-propanol). The wipes which get easily worn out and leave fibers should be avoided. The metallic nets inside the Venturi flow tubes (i.e. tubes with narrow slits for adjusting the volumetric flow) should be cleaned carefully. When dirt settles onto the nets, the flow resistance increases, and consequently the volume flow through the Venturi tubes decreases. This alters the mobility classification. An ultrasonic bath is recommended for cleaning these nets. Instead of nets, the instruments with 3 or 4-blowers have typically honeycomb-shaped pieces to make the flow laminar. These are less likely to become dirty and a careful cleaning with a brush or pressurized air is sufficient. Overall, the 3- and 4-blower instruments are significantly less susceptible to flow deviation issues. The corona-needle chargers should be cleaned (scraped with a sharp knife) regularly (1–3 month intervals) to make sure that the corona-generated ion concentration is maintained at a constant level.

Use plenty of isopropanol and Kimwipes. Do not scratch the inner surfaces of the NAIS. Clean all the surfaces which are in contact with the sample and sheath air flows. Do not wipe the plastic parts with isopropanol, clean them with de-ionized water to avoid leaving a conductive film on the surface. Always wear gloves when handling the DMAs and sample preconditioning units. After the cleaning procedures, check that the ion and particle number size distributions are similar and form a continuous distribution before and after cleaning.

The number concentration of ions and aerosol particles are determined by measuring a current delivered by the flow of charged particles to an electrometer rings. The electrometers are extremely sensitive. The deposition of dirt onto the electrometer ring can deteriorate the signal-to-noise ratio of the electrometer. Dust or fibers that have settled on the electrode may start to form an occasional corona discharge in the electric field of the analyzer. This is the reason why the electrometers facing the bottom inner electrode, which has the highest voltage, are the most likely to become noisy (electrometers no. 13–21). We recommend wiping the electrometer rings clean manually when the average current signal of a certain electrometer increases above few tens of fA (10-15 A) during the offset measurement mode. See Supplement S1 for details.

To clean the electrometer rings without removing and opening the mobility analyzer, remove the top part of the instrument (i.e. sampling preconditioning unit; see Fig. 1). Then use a cleaning rod to clean the mobility analyzer by moving the rod from top to bottom. Figure 10 shows the procedure. Before lifting the top part of the NAIS, ensure that you remove all the nuts holding the plates together and disconnect all the cables coming from the main compartment of the instrument, and disconnect sheath air tubing between the top and bottom part of the NAIS (on both polarities). The top part should be lifted up and placed onto a clean surface, while the open analyzer should be covered to avoid dropping more dirt into it while cleaning. Now the inner and outer electrode of the mobility analyzer (i.e. electrometer rings) can be cleaned with the cleaning rod by moving it up and down inside the analyzer. Avoid scratching the metallic surfaces. The numbering of the electrometers starts from the top to bottom. Remember that the electrometers detecting the smallest charged particles are at the top of the analyzer.

To open the mobility analyzer it needs to be lifted away from its position. The outer electrode of the analyzer should be lifted up, and separated from the inner electrode. Supplement S1 shows, in detail, how to clean the analyzers after opening the mobility analyzer. The electrometers are located on the surface of the outer electrode, whereas the inner electrode has the four voltage sectors to generate the electric field. Take care not to scratch the inner electrode against the outer electrode. After the electrodes are separated, it is possible to clean the electrometer rings by wiping with a clean cloth and some strong solvent like alcohol or isopropanol; see Fig. 11. Wiping should be done by starting from the centre of the electrode and moving towards the top. Take particular care to clean the gaps between the electrometer rings as well as their surfaces. Flip the electrode upside down and repeat the operation to clean the bottom electrometers.

The charging efficiency of the corona charger can change over time as the electrode (i.e. corona needle/wire) gets worn out. For this reason, the NAIS corona needles need to be cleaned regularly or replaced. To clean the corona needles located in the preconditioning unit's charger or discharger, shown in Fig. 12, or in the sheath air filter, shown in Fig. 13, the corresponding parts of the NAIS need to be opened and the needles removed. As shown in Fig. 14, the corona needle can be cleaned from dirt by gently scraping the tip with a sharp knife or by dissolving the dirt. The corona needles break and bend easily so minimum pressure should be applied. When the cleaning does not restore the charging efficiency, the needle needs to be replaced. When replacing the corona needle, be careful and handle the needle with flat tip tweezers. See Supplement S2 for details.

A blower needs to be replaced, when the active feedback loop cannot maintain the right volumetric flows for the sample and sheath flows, which leads to wrong sizing of the aerosol particles. The blower must be sealed properly and a leak test should be done to the instrument before determining the flows. For the 1-blower system, the blower sealing should be done using silicone (e.g. Bostik silicone universal) to seal both the blower itself and to connect it in a leakproof manner to the metal casing, shown in Fig. 15.

The particle diameter is not a well-defined concept at very small sizes or for highly non-spherical particles and agglomerates. We recommend using the electrical mobility equivalent diameter as it can be converted back to the particle electrical mobility, which is the measured quantity by the NAIS. The electrical mobility (Zp) to particle diameter (Dp) conversion should follow the international standard ISO 15900, and use a Millikan–Fuchs equivalent electrical mobility diameter which is based on Stokes' law (Hinds, 1982) Zp=neCcDp3πηDp, where n is number of excess elementary charges e carried by the particle, η is viscosity of air and Cc is the Cunningham slip correction factor for taking account of relation between particle radius and mean free path λ of the gas molecules: Cc(Dp)=1+2λDp1.165+0.483e-0.997Dp2λ. Gas pressure and temperature affect viscosity and the mean free path. Typically the particle size and the mean free path are presented as Knudsen number Kn=2λ/Dp. The constants used in equations follow the ISO15900 standardization as well Kim et al. (2005).

Air viscosity can be written as follows: η=η0TT032T0+110.4KT+110.4K, where η0=1.83245×10-5 kg m-1 s-1. T is the air temperature and T0 is a reference temperature (296.15 K). On the other hand, the mean free path at a reference temperature T0=296.15 K and reference pressure p0=101.325 kPa is λ0=67.3 nm and it can be scaled as follows (Allen and Raabe, 1985): λ=λ0TT02p0pT0+110.4KT+110.4K. The classic Millikan equation may be inaccurate for the finest charged nanometer particles and ions (e.g. Tammet, 1995, and references therein). Mäkelä et al. (1996) analyzed phenomena in detail and compared different mobility equivalent diameters. For the NAIS data, the electrical mobility to electrical mobility equivalent diameter conversion is commonly done following the suggestion of Mäkelä et al. (1996) using the Cunningham slip correction factor with λ=64.5 nm at 296.15 K and 101.325 kPa: CcDp=1+2λDp1.246+0.420e-0.87Dp2λ. This approach is in excellent agreement with the ISO standard in normal temperature and pressure (NTP) conditions (293.15 K, 1013.25 hPa). Note that the geometric diameter (i.e. Tammet's mass diameter; Tammet, 1995), which is related to particle mass, is about 0.3 nm smaller than the electrical mobility diameter (Mäkelä et al., 1996; Ehn et al., 2011).

The raw signal of the NAIS is calculated from the current I (A) measured by one of the NAIS electrometers, which is related to the sampled aerosol concentration N (ions cm-3) as follows: I=NnpeQs, where np is the average number of elementary charge units per particle (assumed to be unity), e=1.6×10-19C is the elementary unit of charge, and Qs (cm3 s-1) is the volumetric sample flow rate passing the electrometer. When the average number of elementary charge units per particle is known, the aerosol concentration can be calculated from Eq. (6).

The NAIS measures all electrometer signals approximately 12 times per second. Data is averaged typically over 1 s, 10 s and measurement cycle (block) periods. The instrument measures a large number of secondary parameters that describe the detailed state of the whole system. The average electrometer signals together with the secondary measurement parameters are stored into record files by the Spectops measurement software.

The average electrometer signals are converted into ion mobility or particle size distributions by the Spectops software and stored in spectra files. The distribution is calculated using the generalized least squares method to find the size or mobility distribution that best matches the measured electrometer currents according to the instrument matrix while taking into account the noise level estimates. The instrument matrix is based on a mathematical model of the instrument that considers particle losses, charging probability (in case of particle mode measurements), electric field and air flow inside the mobility analyzer.

It is important that the user always stores the record files together with the spectra files containing the measured distributions. The secondary measurement parameters stored in the record files are vital for confirming the validity of the measurements and for problem diagnostics with the instrument. The spectra files can be recalculated from the electrometer signals stored in the record files as part of pre-processing and data analysis.

The sampled particles are assumed to be in charge equilibrium. The particle charging probability is predicted by Fuchs' diffusion charging theory (Fuchs and Sutugin, 1971). At a constant corona-wire current, the aerosol charging depends mainly on the particle size, on the charger ion concentration and on the residence time of the aerosol in the charging region. The product of the latter two is called nt product. The model that performs the NAIS inversion, takes into account measured aerosol volumetric flow rates, particle charging probabilities, size dependent loss factors, and the charging parameter (i.e. nt product). The charging probabilities use a calibrated charging parameter α=6, which translates to nt =2.22×106 for a mobility of 1.5 cm2 V-1 s-1. According to these numbers, 1 % of 1 nm particles and 5 % of 5 nm particles are singly charged.

Although the data inversion assumes that the sampled particles are in a charge equilibrium, the unipolar charger does not neutralize the aerosol sample entering the NAIS (McMurry et al., 2009). Thus, if the sample is highly overcharged, this can lead to an overestimation of the ion and particle concentrations.

In the ion mode, the inversion considers only charged particle mobility. It does not make assumptions about their charging probability or background aerosols. Therefore, it produces a mobility distribution which the user will later convert into a size distribution assuming that all the detected ions are singly charged. To simplify, in ion mode we make an assumption that all charged particles are singly charged. In practice, this means that ion concentrations in the size range from ∼ 20 to 40 nm may be overestimated and the shape of the distribution may be distorted as a part of these particles carry more than one charge.

Moreover, Alguacil and Alonso (2006) reported that when using a corona discharge, a substantial fraction of doubly charged particles occur in the particle diameters down to ∼ 15 nm. Therefore, the measurement uncertainty of the NAIS increases above 20 nm because the corresponding electrometers are also affected by the multiply charged particles with diameters up to 90 nm. Due to the limited mobility the range of the DMA, the data inversion cannot completely account for these effects stemming from larger particles. Therefore, we recommend that the ion and particle number size distributions above 20 nm should be utilized with caution. One possibility is to merge the number size distribution measured with the NAIS into a distribution measured with, e.g. a differential mobility particle sizer (DMPS, Wiedensohler et al., 2012) in size range from 10 to 1000 nm to obtain information on the background aerosol population to be used in the data inversion (e.g. Kulmala et al., 2012).

A detailed description of the mathematical model of the NAIS is presented in Mirme and Mirme (2013). The instrument response of the NAIS is a set of electric currents that are generated by the flux of ions precipitating on the collecting electrodes. An ion mobility distribution f(z) is linked with the electrometer currents yi (i=1…n) using the analyzer response function G(i,z) as follows: yi=∫eG(i,z)f(z)dzi=1,…,n, where e is the elementary charge.

The analyzer transfer function G(i,z) is the response of an electrometer i to a singly charged ion with a mobility of z. The function is derived in a straightforward way presented in Mirme and Mirme (2013) and it is verified by calibration measurements in the ions mode. Diffusion losses for the sampled particles inside the instrument are estimated by fitting a diffusion length parameter of the instrument model to the ion mode calibration results. The diffusion and electrostatic losses in the sampling lines prior to the instrument should be taken into account by the user.

During particles measurements, the corona charger is active. The instrument response in the particles mode includes the charging probability functionP(q,r): yi=∫e∑qqG(i,z(r,q))P(q,r)f(r)dr(i=1,…,n), where f(r) is the particle size distribution andP(q,r) is the probability that a particle with a radius of r carries q elementary charges.

The analyzer transfer function G(i,z) is identical for both ions and particles measurements. The charging probability function P(q,r) is based on a theoretical charging model. The function is adjusted and verified using calibration measurements in the particles mode.

The data inversion finds an approximate ion mobility distribution f(z) or particle size distribution f(r) that best satisfies Eq. (7) or (8). The distributions are estimated as a sum of predefined elementary distributions Fi(r). f(r)=∑jϕi⋅Fi(r). This allows to transform Eq. (7) or (8) to an equivalent matrix form as follows: yi=∑jHij⋅ϕj(i=1,…,n), where H is the instrument matrix and Hij determines the response of the electrometer i to the predefined distribution Fi(r). The data inversion procedure solves the matrix Eq. (10) for the spectrum vector ϕj and calculates the size or mobility distribution estimate using Eq. (6).

The ratio of sample flow to total analyzer flow is about 1 : 3 for the NAIS, which is quite large and therefore even perfectly monomobile particles will have a response on several electrometers. For the particles with diameters above 20 nm the probability of acquiring more than one elementary charge in the corona charger is non-negligible. Hence the electrometer response for the larger particles becomes even wider. This also means that the measured electrometer signal may be a combination of by both singly charged smaller particles and multiply charged larger particles with the same mobility.

The multiply charged particles do not require special treatment in the data inversion. They are naturally included in the calculated response of the electrometers, i.e. the instrument matrix. However, the uncertainty of the measurement results gradually increases for particle sizes above 20 nm because the electrometer responses become less distinguishable for the larger particles and because a gradually larger portion of the response will be lost beyond the mobility range of the NAIS.

The electrostatic losses inside the corona charger during charging process lead to underestimation of the particle concentration (Alonso et al., 2006; Huang and Alonso, 2011). The diffusion losses decrease and electrostatic loss increase as the charger voltage is increased, whereas charging efficiency increases with particle size and charger voltage. Electrostatic loss of small particles increases with decreasing particle diameter. The electrostatic losses in the sampling lines prior to the instrument should be taken into account by the user.

After the data collection and prior to the data analysis, the data should be quality controlled. The bad data should be removed from the final data. Figure 16 illustrates some examples of typical faulty spectra and the data needed to be removed. Data quality checks and criteria, which should be fulfilled for the ion number size distributions, are the following: (1) negative and positive ion number size distribution agree visually (a similar distinct shape for number size distributions in both polarities), (2) size distribution has a continuous cluster ion mode visible in both polarities with a mode peak at ∼ 1 nm; see Fig. 16b, (3) when looking at the negative ion size distribution, the mean diameter of cluster ion mode should be slightly smaller compared to positive cluster ions (typically one channel difference between the peaks for the polarities), (4) in the number size distribution plots, the smallest of Aitken mode charged particles should be visible and have a diurnal variation at 25–42 nm size range, and (5) time series of total ion number concentration between polarities should agree within 10–20 %, which should apply to small, intermediate ions and large ion concentrations as well, and (6) when intermediate ions are observed, make sure that the size distribution does not have any gaps due to instrument malfunction; see Fig. 16d.

The primary data quality checks for the particle number size distributions are similar as for the ion data. In addition, check that the corona-charger-generated ions do not dominate the particle spectra due to inadequate post-filter settings: (1) the 2–3 nm particle concentrations should remain in a range of ∼ 200–700 cm-3, when no new particle formation event is taking place to maintain optimal post-filter setting, (2) determine the smallest detectable size using both positive and negative polarities (see Sect. 3.7.4), and (3) select the preferred polarity to give the final data for measured particle number size distribution (Sect. 3.7.5).

If possible the ion and particle data measured with the NAIS should be crosschecked with the data from additional instruments. The data quality check for the offset mode: electrometer currents during offset measurements should not exceed ±10 fA.

In the ion mode, the Spectops inversion algorithm converts signal from 21 electrometer signals into 28 normalized ion mobility distributions, dN/d(log⁡10Zp), which the user has to convert into ion size distributions, dN/d(log⁡10Dp). We recommend doing the mobility to diameter conversion using the Millikan–Fuchs equivalent mobility diameters introduced in Sect. 3.6.1.

To do this, the user should follow steps: (1) Open the spectra data files to get the geometric means of all 28 mobility fractions and calculate the lower and upper mobility limits for each mobility fraction. (2) Calculate the d(log⁡10Zp) for all the mobility fractions using these limits. (3) Calculate the absolute number concentration for each mobility fraction starting from the normalized ion mobility fractions: (dN/d(log⁡10Zp)×d(log⁡10Zp)=dN. (4) Calculate the corresponding mobility diameters for each mobility fraction limits by determining the lower and upper limiting diameters, and calculate the d(log⁡10Dp) for all size fractions. (5) Normalize the absolute concentrations using the determined size fractions: dN×(1/d(log⁡10Dp))=dN/d(log⁡10Dp).

Diffusion losses inside the sampling lines prior to the instrument; see Fig. 7 (upper panel), should be taken into account by post-processing of the data. The particle losses by diffusion in a straight inlet line can be described by calculating a size-dependent particle penetration (Hinds, 1982). In laminar flow, these losses depend only on the line length, the flow rate through the line, and the particle size. In cases that bends cannot be avoided in the sampling pipe, the size-depended particle penetration can be calculated according to Wang et al. (2002).

In the particle mode, the raw data is typically converted into 29 particle size distributions, dN/d(log⁡10Dp), by the Spectops software. During the particle mode measurements, the corona-generated ions complicate the particle detection as the measurement size range overlaps the size range of the charger ions. The positive and negative corona-generated ions are smaller than 1.8 and 1.6 nm, respectively, which results in the lower detection limit of approximately 2 nm for the NAIS particle measurements (Manninen et al., 2011). Therefore, the particles below about 2 nm cannot be reliably distinguished from the corona-generated ions. Typically, the lowest detection limit for the NAIS in the particle mode is between 2 and 3 nm depending on the corona voltage and on the properties and composition of carrier gas (environmental conditions). It is important to notice that the lowest detection limit of the NAIS varies according to the environmental conditions (Manninen et al., 2011). The lowest detectable size of the NAIS in particle mode should be checked on regular basis, at least separately for different seasons.

During subsequent data processing the corona-generated ions below the lowest detection limit should be always cut out of the particle number size distribution. The lowest detection of the NAIS is equal to the upper edge of the corona ion size distribution which is illustrated with the gray shaded area in Fig. 3. The determination of the lower detection limit should be done always when no new particle formation (i.e. natural cluster activation and growth) is taking place. If possible, after removing the corona ions from the particle number size distribution, the particle concentrations in the 2-3 nm range should remain in a concentration range of ∼ 200–700 cm-3 to make sure that not all charged particles are cut out together with the corona-generated ions.

Due to the design of the instrument, the particle spectra (i.e. particle number size distribution) is measured with both positive and negative corona charging. Ideally the distributions should be identical. A large and a persistent difference may indicate a problem with the measurements. We recommend giving only one particle spectra in the final processed data to avoid misunderstanding. Thus, the user needs to decide on the preferred polarity on the particle data, which is reported as the final particle number size distribution. Typically, the preferred polarity is chosen with two main criteria: (1) the polarity which has lower background level of the corona-charger ions extending to 2–3 nm size range (which is considered the lowest detection limit of the NAIS in particle mode); (2) the polarity, which shows no short-term fluctuation over time (i.e. corona-charger ion background stays same level over a diurnal cycle). (3) The small corona-charger ion background (∼ 102) should be visible at 2–3 nm to be sure that the particles in the aerosol sample are not filtered with the electric post-filter.

Charged particles are divided into small ions (1.3–0.5 cm2 V-1 s-1), intermediate ions (0.5–0.034 cm2 V-1 s-1), and large ions (0.034–0.0042 cm2 V-1 s-1), which correspond to mobility diameters of 0.8–2, 2–7 and 7–20 nm, respectively (Hõrrak et al., 2001). As can be seen from Fig. 17, the ion number size distribution and time series of ion concentration measured with the NAIS has a distinct shape during new particle formation. The concentration of the small air ions in the atmosphere is determined by competition between production and loss processes (e.g. Israël, 1970; Hirsikko et al., 2011). The small ions i.e. cluster ions are detected in all environmental conditions, where they have been measured with the ion spectrometer varying from extremely polluted areas (e.g. Backman et al., 2012; Herrmann et al., 2014) to extremely clean environments (Virkkula et al., 2007) as well as from the lower troposphere to the free troposphere (e.g. Manninen et al., 2010; Mirme et al., 2010; Rose et al., 2015). The only exception where no cluster ions were observed is when measured inside a cloud (Lihavainen et al., 2007) or during a rapid, extreme increase in particle number concentration (Jayaratne et al., 2015). On the other hand, the intermediate ions are a strong indicator for the secondary aerosol formation in various environments (Dos Santos et al., 2015; Leino et al., 2016), whereas the large ions represent the naturally charged fraction of Aitken and accumulation mode particles.

Ionization of air molecules (e.g. N2 and O2) produces primary air ions: positive ions and free electrons. The primary ions undergo rapid chemical reactions, getting neutralized and charged again, and become small ions in less than a second from their formation (Hõrrak, 2001). To avoid misunderstandings, it should be noted that these primary ions are not detected by the NAIS as their electrical mobility is too high to be classified with the NAIS and their lifetime is too short for the detection with the current version of the instrument.

The lowest detection limit for the NAIS in the particle mode is approximately 2 nm due to overlapping corona-charger ions (Asmi et al., 2009; Manninen et al., 2011). Thus, the NAIS is not able to detect the pool of stable neutral clusters at sub-2 nm (Kulmala et al., 2013). The NAIS can detect only a “shoulder” of this neutral cluster pool. The NAIS in a very capable tool for detecting the newly formed particle already at the 2–3 nm size range depending on the post-filter settings. As an example, the time series of 2–3 and 3–6 nm particle concentrations on new particle formation day are shown in Fig. 18. In the particle mode, the NAIS overestimates the total particle number concentrations by a factor of 2–4 (Manninen et al., 2009; Gagné et al., 2011). The quantitative agreement improves at conditions representing particle formation bursts when higher particle concentrations are typically observed in the overlapping size range (nucleation mode). As seen in Fig. 19, merging particle number size distributions measured with the NAIS (2.5–40 nm) and a differential mobility particle sizer (DMPS, 40–1000 nm) without any additional fitting highlights the problem at 20–40 nm range where the agreement is poor, as the NAIS overestimates particle concentrations. Therefore, we recommend using the particle spectra measured with the NAIS up to 20 nm.

Neutral particle formation seems to dominate over ion-induced and ion-recombined nucleation, at least in the continental boundary layer (Lovejoy et al., 2004; Manninen et al., 2009, 2010; Zhang et al., 2012; Kulmala et al., 2013). The results obtained from the NAIS particle and ion measurements agree well with separate independent measurements performed with other electrical mobility spectrometer (e.g. Gagné et al., 2011) and condensation-based techniques (Lehtipalo et al., 2009, 2010; Kulmala et al., 2013; Rose et al., 2015). The atmospheric ions participate in the initial steps of the new particle formation, although their contribution has been shown to be minor in the boundary layer (e.g. Kulmala et al., 2013). The highest atmospheric particle formation rates are observed at the most polluted sites, where the role of ions was the least pronounced (Manninen et al., 2010). Furthermore, an increase of particle growth rate with size suggests that enhancement of the growth by ions is negligible (Yli-Juuti et al., 2011).

By conducting measurements according to the work presented here, it is possible to develop simple yet sufficiently accurate nucleation parameterizations for large-scale atmospheric modelling. The secondary aerosol formation includes the production of nanometer-sized clusters from atmospheric vapours and the growth of these clusters to larger particles. One dynamic process modifying the size distributions of neutral and charged clusters is ion–ion recombination, which was parameterized by Kontkanen et al. (2013). Nieminen et al. (2011) derived a parameterization for the ion-induced nucleation or, more precisely, for the formation rate of charged 2 nm particles. In addition, it is important to predict nanoparticle growth accurately in order to reliably estimate the atmospheric cloud condensation nuclei concentrations. Häkkinen et al. (2013) introduced a semi-empirical parameterization for sub-20 nm particle growth that distributes secondary organics to the nanoparticles according to their size and is therefore able to reproduce particle growth observed in the atmosphere. All semi-empirical parameterizations described here are based on extensive NAIS datasets that enable to test how well the parameterization captures the seasonal cycle of the modelled parameters and to determine the required weighing factors in different environments. Leppä et al. (2009) introduced an aerosol dynamical box model, which includes basic dynamical processes (e.g. condensation, coagulation and losses by deposition) as well as ion–aerosol attachment and ion–ion recombination. This model was validated and constrained against the NAIS data.

Small ions are almost always present in the air and are responsible for the atmospheric electrical conductivity (e.g. Harrison and Carslaw, 2003). The early research of air ions was mainly focused on atmospheric electricity to study, e.g. air quality (Israël, 1970). Tammet et al. (2009) suggested that the air ions and the atmospheric electric field controlling the migration of ions should be considered, when discussing the formation of primary and secondary particles. Air (polar) conductivity can be calculated directly from the ion number size distributions measured by the NAIS, in addition to reporting the concentration of small, intermediate, and large ions, and the average small ion mobility.