Introduction

The rapidly growing nanotechnology market is a cause for research and discussion in public and occupational health due to the numerous applications of nanomaterials in many industrial sectors [1]. These are associated with an increased likelihood of exposure in the working environment, which can have an adverse effect on human health or the health of living organisms in general [2]. Nanomaterials are characterised by specific physicochemical properties due to their structure, size, shape and surface modifications, usually different from their bulk counterparts, so the risks they may pose to human health are new and not yet fully understood [2,3,4]. Nanomaterials most often penetrate organisms through the respiratory system, causing adverse health effects at this point of entry, such as local inflammation. They can enter the cardiovascular system from the lungs. Through the bloodstream, they have the potential to reach specific organs such as the liver, spleen, kidney, heart, brain, lymph glands, or cells of the reproductive system [4,5,6,7]. Certain types of nanomaterials have been assigned by the International Agency for Research of Cancer (IARC) to Group 2B (possibly carcinogenic to humans): Mitsui-7 multiwalled carbon nanotubes (MWCNT-7), carbon black and nano-TiO2 [8, 9]. Nanomaterials are also supposed to share the same biological mechanisms of health adverse effects, with ambient ultrafine particles (UFPs), therefore the knowledge from one field can be translated into another [10]. To better understand the mentioned effects, there is a need for future research on the toxicity of nanomaterials and further development of nanotoxicology [2, 3, 11].

As nano-objects may pose a risk to human health, innovations in nanotechnology should be accompanied by achievements in the monitoring of nanomaterials. For characterisation, direct and indirect measurements of number, mass and surface area concentration are currently most commonly used. However, nanoparticles (NPs), when considering particles in air, unlike microaerosols, have a much greater impact on number concentration than mass concentration due to their small size. Therefore, measuring number concentration is a suggested proposal for characterising NP exposure [12,13,14]. Currently, the most commonly used devices for NP characterisation are Scanning Mobility Particle Sizer (SMPS), Electrical Low-Pressure Impactor (ELPI), Diffusion Size Classifier (DSC) and Condensation Particle Counter (CPC). The principles of measurements of these devices are further explained in several other papers [13, 15,16,17]. These devices for determining NP concentrations are mainly used in research laboratories and may not be the best choice that helps monitor the NP concentrations in a large production facility, especially due to the heavy weight and costs, i.e. purchase, conservation and maintenance of equipment [15,16,17,18]. Therefore, there is a growing demand for monitoring devices suitable for similar applications as laboratory-grade devices and with good performance. This is especially true in terms of NPs’ concentration in the workplaces (nanotechnology sector), research institutes and universities.

Equipment capable of monitoring the concentration of NPs, that is affordable to be installed for continuous monitoring in areas where such particles are used or that may be formed spontaneously, is now in the field of interest. The use of ionisation chambers, similar to those used in smoke detectors, to measure particle concentrations has already been the subject of several studies [19,20,21]. To measure both submicrometric and micrometric aerosols, Litton et al. [19] have proposed the use of a combination of an ionisation chamber and an angular scattering sensor. Edwards et al. [20] showed that their particle monitor is characterised by good linearity in response to the different concentrations of used aerosols, and the ionisation chamber was about five times more sensitive when examining fine particles than coarse ones. Dahl et al. [21] have proposed a method for low-cost NP concentration monitoring using a modified ionisation smoke detector. To avoid the influence of environmental conditions on the results, they measured the difference between filtered and untreated air. The detection limit of 100 nm particles was 15,000 particles/cm3. The tested sensor was calibrated with sodium chloride (NaCl) aerosol and tested with candle smoke and welding fumes.

This paper presents the results of research on application tests of developed ionisation nanoparticle detectors, whose principle of operation is based on a modified ionisation-type smoke detector (same as for detecting fires at an early stage). A study on the use of the ionisation sensor conducted in the Laboratory of Nanoaerosols in the Central Institute for Labour Protection – National Research Institute confirmed that it can be used to measure the number concentration of nanoaerosols [22, 23]. In the first test, it has been shown that the sensor output changes linearly with the change of diesel soot concentration in the range up to 8.3 × 105 particles/cm3 [22]. The next research confirmed that an ionization detector could be used to measure the concentrations of nanoaerosols [23]. The modified smoke detector detected soot particles smaller than 100 nm. The aim of the following study was to determine the effect of the number concentration of NPs on the analogue output signal from a modified smoke detector and to define the correlation equation. Detector tests were done in two stages: with and without aerosol neutralisation. Studies of the response of the ionisation detector to changes in NP concentrations were carried out under controlled laboratory conditions.

Materials and methods

The main part of the developed ionisation nanoparticle detector, presented in Fig. 1, responsible for the measurement was the modified ionisation smoke detector (DIO-40, Polon-Alfa S.A., Poland) [24]. It contained a radioactive source (241Am) with an activity of 7.4 kBq which generated α particles to cause an ionisation of the air in the two connected chambers. Both positive and negative ions were formed and present in the vicinity of the radiation source. The rest of the chamber contained only one type of ion, attracted by the electric field. Between the two chambers, a space-charge region was developed, which acted as a boundary layer [25]. When visible and invisible aerosol particles entered the chamber and ions attached to them, there was a decrease in the ionisation current in the measuring chamber, normally in equilibrium with the second chamber, due to a decrease in the mobility of the carriers — ions. Visible particles are those with a size in or above the visible spectrum, i.e. size in the range 380–750 nm or above. However, in pharma sciences, particles visible to an unaided human eye are defined as particles of size greater than or equal to 100 μm [26].

Fig. 1
figure 1

Block diagram and view of ionisation nanoparticle detector

Full size image

Two signals from the smoke detector were exposed: supply voltage and a signal from the electrometer connected to the floating electrode. The difference between these two analogue signals was measured using an electronic circuit. First, both signals were amplified with a double operational amplifier (LM358, Texas Instruments Inc., USA) with a gain equal to 1. Secondly, to eliminate manual selection of resistors with possibly similar resistance values, which would have influenced the precision of the measurements carried out, an integrated circuit (AMP03GSZ, Analog Devices Inc., USA) was used to determine the difference between the two signals. The value then was brought below the reference voltage using a resistor divider. The signal thus obtained was received using an analogue-to-digital converter (ADC) built into a Microchip SAMD21 microcontroller (ItsyBitsy M0, Adafruit Industries, USA). The resolution of ADC was set to 12 bits, and the reference voltage from the board was used. Code was developed in MicroPython for the Adafruit ItsyBitsy M0 board, which read the output signal from the modified smoke detector with an interval of 1 s and transferred to the main microcontroller (Raspberry Pi Pico, Raspberry Pi Ltd, UK) for further calculations.

The device consisted of two air ducts: the first for air filtered through a high-efficiency particulate air (HEPA) filter (9933-05-CQ, Parker Hannifin Corp., USA) (reference level) and the second for air containing nano-objects (untreated air). The two paths were separated by an electrovalve (2N08-SS Viton 1/4”, HPCONTROL SEBASTIAN BÓJCZUK S.K.A., Poland) that opened and closed during the measurement, allowing air to flow through the first duct (3.6 l/min) and the second duct (5.9 l/min). The principle of the ionisation nanoparticle detector was based on the fact that, in one measuring cycle, it measured the output signal from the modified smoke detector, which was the difference between the supply voltage and the output voltage from the measuring chamber, once for filtered air and later for untreated air. The difference between these signals (further named as the difference in the output signals) for filtered air and air containing nano-objects was then determined in order to isolate the effect of aerosol particles alone on the signal value, among other parameters that could have influenced the signal, such as environmental conditions or air velocity. These signals were measured in still air, when the vacuum pump (ROB-10398, SparkFun Electronics, USA) was switched off. Using filtered air as a reference has two main advantages:

  • no need for calibration

  • no impact of pollution accumulation in the chamber.

The test stand located in a sealed test chamber is shown in Fig. 2. The test chamber with a working volume of approximately 1 m3 was purged with filtered air (HEPA filter) before the start of each measurement. The chamber was equipped with tubing ports with different diameters (8, 10, and 12 mm), two of which were used to connect an aerosol generator outlet and a reference device. The experiment was conducted in two stages: with and without neutralisation of the aerosol before it was fed into the test chamber.

Fig. 2
figure 2

The test stand with ionisation nanoparticle detector

Full size image

NPs used for testing were generated by a nanoaerosol generator (GFG 1000, Palas GmbH, Germany) with graphite electrodes. The device generates carbon particles in an argon stream, whose size distribution and morphology can be varied by adjusting the spark frequency. For the purpose of this study, spark frequency was set to 6 Hz ensuring that the concentration and particle size distribution remain constant [27]. GFG 1000 was located outside the test chamber and the aerosol was fed into the chamber via a flexible antistatic polyurethane tube (1025U08A01, Parker Hannifin Corp., USA) at a flow rate of 5.2 l/min. During the measurements, the conditions were changed, i.e. the aerosol generator was switched on and off, in order to study the response of the equipment to varying values of the number concentration of nano-objects in the test chamber. Carbon particle aerosol generation lasted between 3 and 5 min in each case. According to the available data, GFG 1000 as a spark discharge generator, generated positively-, negatively- and non-charged particles with the latter predominating. For metal particles (aluminium, copper, silver and gold), non-charged particles should have accounted for approximately 90%, and in some cases, almost 100% of all generated particles [28, 29]. However, for carbon particles, the ratio should have been lower, decreasing with size from 90 to 55% [28, 30]. The rest would have been charged particles with a predominance of negatively charged particles, due to higher mobilities of negative ions than positive ones [29, 31]. The charge accumulated on the particles could have been as high as a few elementary units of charge [29, 30]. However, Simones illustrated in his thesis a slight predominance of positively charged particles. As the author pointed out, it would have been better to use a neutraliser to confirm or disprove the ability of the tandem differential mobility analyser (TDMA) used to measure the expected equilibrium distributions [30]. Neutralisers bring particles into equilibrium charge distribution according to the theory by Fuchs [32, 33] which enables calculation for all particles. Hence, the motivation for the current research was to check whether the neutralisation of the generated particles would affect the results obtained by the ionization detector. Charged particles bring a certain charge to the measuring chamber, which may affect the measured potential of the floating electrode, and in the case of our detector, this is a component of the output signal. The result of the work is to check the influence of the use of a charge neutraliser on the output signal from the detector and to compare the results obtained for neutralised and non-neutralised aerosol.

As a reference device, an aerosol spectrometer (MINI-WRAS 1.371, Grimm Aerosol Technik GmbH & Co. KG, Germany) was used due to the wide test particle size range — from 10 nm to 35 μm. The MINI-WRAS 1.371 combines two measurement methods: particle detection based on electrometer readings (particle size from 10 to 193 nm, i.e. electrical mobility diameter) and measuring scattered light intensity (particle size from 0.253 to 35 μm, i.e. optical equivalent diameter). The particle diameter values shown in brackets refer to the lower limits of the size classes. The combination of the two measurement methods makes it possible to determine the size distribution of aerosol particles over 41 size classes within the measuring range of the instrument. A single measurement takes 1 min. MINI-WRAS 1.371 detection limit declared by the manufacturer for particles measured with the electrometer is 3000 particles/cm3. However, for the laser spectrometer, the manufacturer declares that the device is capable of determining concentrations from 1 particle/l. Purging the chamber meant such a concentration of particles that was well below the MINI-WRAS 1.371 electrometer detection limit (to ensure, once the limit was reached, purging lasted another 10 min). MINI-WRAS 1.371 itself was located outside the test chamber but its inlet was extended with a flexible tube to place in the chamber near the impactor of the ionisation nanoparticle detector, at a distance of approximately 40 cm from the port, into which the aerosol stream generated by the GFG 1000 was fed.

In the second stage for the aerosol neutralisation a neutraliser with radioactive source 85Kr with an activity of 740 MBq (Model 3054A, TSI Inc., USA) was used. The neutraliser was placed outside the chamber in the path of the generated aerosol, before the generated aerosol stream was fed into the test chamber at a flow rate of 5.2 l/min. Therefore, the aerosol was neutralised before being fed into the chamber. At this stage, air parameters were also tested in the chamber. For this purpose, thermo-hygro-barometer (LB-706B, LAB-EL Laboratory Electronics, Poland) was used.

The sampling time, i.e. the time taken for the microcontroller to read the output signal from the modified smoke detector in the ionisation nanoparticle detector during all tests, was constant and equal to 1 s. During the tests, the averaging time of the signals received by the detector was changed. In the first stage, tests were conducted at a single-fixed detector averaging time of 10 s. However, in the second stage, tests were conducted at different detector averaging times. The tests were carried out in the test chamber for more than 2 h of continuous measurements, during which measurement data were collected to determine a correlation between the number concentration of the generated aerosol particles and the difference in the output signals of the ionisation nanoparticle detector.

In order to test whether there were statistically significant differences between measurements done during first and second stage, i.e. without and with neutraliser, an independent sample t-test was used. When the data did not meet the assumptions of normality (tested with the Shapiro-Wilk test) and equality of variances (tested with the Levene’s test), a Mann–Whitney U test was used instead of an independent sample t-test. Basic descriptive statistics and a correlation series with Pearson’s r coefficient were also determined. Significance was considered to be p < 0.05. Statistical analysis was performed using Python (version 3.9.15), libraries statsmodels (version 0.13.5) and SciPy (version 1.9.3).

Results

Test particles

Size distributions of carbon particles generated by GFG 1000 in two stages are shown in Fig. 3. The figure depict the distributions combined by the instrument acquired from the optical aerosol spectrometer and the electrometer across the aforementioned size ranges. In the first stage, the particles in the test chamber were characterised with a number-weighted mean particle diameter of 63 nm. Particles with a size smaller than 100 nm accounted for 94.85% of all particles. In the second stage, the number-weighted mean particle diameters were 71 nm for detector averaging times of 10 and 15 s and 70 nm for 20 s. Particles with a size smaller than 100 nm accounted for 90.35% and 91.39% of all particles, respectively. In both the first stage and the second stage, particles detected by the electrometer accounted for 99.97% of all particles. Hence, it can be reasonably inferred that Fig. 3 illustrates the size distributions in relation to the electrical mobility diameter. This may also mean that the detector is particularly suitable for detecting non-neutralised nanoparticles.

Fig. 3
figure 3

Averaged particle size distribution of the generated aerosol during tests in the test chamber: a first stage, b second stage for averaging times of 10 and 15 s, and c second stage for averaging time of 20 s

Full size image

First stage (without aerosol neutralisation)

In the first step, basic descriptive statistics were calculated, as shown in the Table 1. Points below the MINI-WRAS 1.371 detection limit were removed from the dataset. The obtained values of the number concentration of particles generated by the GFG 1000 was high, reaching almost 350,000 particles/cm3. Shapiro-Wilk tests were performed to determine the conformity of distributions with the Gauss curve with 95% confidence – p-value in Table 1. Figure 4 presents the obtained correlations for selected time intervals, taking into account the largest number of results recorded by both devices, with a shaded 95% confidence region. Linear model coefficients, standard errors and their 95% confidence intervals (CI) are shown in the Table 2.

Table 1 Descriptive statistics — first stage
Full size table
Fig. 4
figure 4

Correlations between the number concentration of NPs and the difference in output signals for the time intervals a 12 min and b 44 min

Full size image
Table 2 Linear model results — first stage
Full size table

Tests of the ionisation nanoparticle detector in the test chamber without aerosol neutralisation, carried out using a spark generator with graphite electrodes, made it possible to determine the relationship between the number concentration of particles and the difference in the detector output signals in several measurement series. Correlations were strong, Pearson correlation coefficient r was 0.855 and 0.880 for Fig. 4a and b, respectively, and was linear with the slope and the y-intercept of similar values for the two cases considered.

Second stage (with aerosol neutralisation)

The tests were conducted at three different detector averaging times: 10, 15 and 20 s. Air parameters were also recorded at this stage and their changes are shown in Figs. 5 and 6. The temperature rose successively during the measurement from about 19 °C to about 21 °C, with temporary decreases occurring when the chamber was purged between each measurement (see Fig. 6). The relative humidity, which rose successively from about 31 to about 35% during the measurement, also increased during the cleaning of the chamber. The ambient pressure successively decreased during the measurement from 101.0 to 100.6 kPa.

Fig. 5
figure 5

The time-course variation of temperature and relative humidity during the tests in the test chamber (second stage)

Full size image
Fig. 6
figure 6

The time-course variation of ambient pressure during tests in the test chamber (second stage)

Full size image

As in the first stage, the basic descriptive statistics, shown in Table 3, were calculated first, removing points below the MINI-WRAS 1.371 detection limit from the dataset beforehand. The obtained values of the number concentration of particles generated by the GFG 1000 were even higher than in the first stage, exceeding even 400,000 particles/cm3. Shapiro-Wilk tests were performed and the p-value was shown in Table 3.

Table 3 Descriptive statistics — second stage
Full size table

It was decided to test the ionisation nanoparticle detector in the test chamber with aerosol neutralisation due to the operating characteristics of the spark generator used. A charge that could have been present on the generated particles could have affected the magnitude of the detector signal since the output signal was a voltage. The generated aerosol was characterised by similar parameters as in the first stage, although in this case even higher number concentrations of particles in the chamber were achieved, exceeding 400,000 particles/cm3. The correlations obtained in this case, shown in Fig. 7, were also strong and linear with slope and y-intercept of similar values for the three cases considered, although obtained for different averaging times (10, 15 and 20 s). Pearson’s correlation coefficient r was 0.942, 0.903 and 0.956 for Fig. 7a, b and c, respectively. Linear model coefficients, standard errors and their 95% confidence intervals (CI) are shown in Table 4.

Fig. 7
figure 7

Correlations between the number concentration of NPs and the difference in output signals for the time intervals a 19 min, averaging time: 10 s; b 50 min, averaging time: 15 s; and c 19 min, averaging time: 20 s

Full size image
Table 4 Linear model results — second stage
Full size table

During these measurements, there were no significant changes in the environmental conditions that could have significantly altered the results obtained, for two reasons: measurements were done in 30–40 min intervals when the change of air parameters was negligible and the principle of the detector itself, which was comparing signals for filtered and untreated air in order to minimise the influence of these parameters on the measurements. What is more, the largest fluctuations in these parameters occurred when the test chamber was purged, when measurements of the number concentration of particles and difference in the detector output signals were not carried out.

Discussion

The results of the Shapiro-Wilk tests for introduced variables proved to be statistically significant (p < 0.05) in each case, meaning that their distributions significantly deviated from the normal distribution. It should have been noted, that the skewness of the distribution of variables was also negligible in each case, meaning that their distributions were asymmetric to a negligible degree [34]. Non-normality of distributions of measured variables was expected due to the nature of the experiments and data, which were characterised by specific concentrations change in a predetermined manner rather than random variation. The instruments displaying randomly fluctuating values instead of the specifically measured quantity would have not been useful devices.

For further statistical analysis, Levene’s test was performed to check the equality of variances of the corresponding data sets from both stages. The results obtained from the test indicated that there was homogeneity of variance in the distributions of the measured variable — number concentration of particles — from both stages (p = 6.340 × 10−1), and heterogeneity of variance of the measured variable — difference in output signals — from both stages (p = 4.224 × 10−2). Considering the above and the Shapiro-Wilk test results, the Mann–Whitney U test must have been conducted to compare distributions of corresponding variables from both stages. The results indicated that there was not a statistically significant difference between the number concentration of particle distributions measured with MINI-WRAS in both stages (p = 7.182 × 10−2); however, there was a statistically significant difference between the difference in output signals distributions from both stages (p = 4.519 × 10−7). Specifically, these results indicated on the similarity of conditions for the generated aerosol, because MINI-WRAS acted as a reference device. However, for the investigated ionisation nanoparticle detector, the influence of aerosol neutralisation on the measurement results was seen. The charge accumulated on the particles may have influenced the measurement data obtained, which was also seen in correlations. Correlation equations obtained from first-stage data differed from those obtained in the second stage. Relative to the neutralised aerosol, the detector was more sensitive (the correlation equation obtained was characterised by a larger slope) and the measuring range of the instrument was shifted (the y-intercept in absolute value was larger).

During measurements below a certain value, an increase in the scatter of the results obtained from the detector was observed in each case. Therefore, a lower limit of 45 mV was proposed for the difference in the detector output signals. Taking into account all the measurements carried out without aerosol neutralisation and the lower limits of the measuring ranges of the equipment (for MINI-WRAS — 3000 particles/cm3, for detector — 45 mV), a correlation between the number concentration of particles (measurement results from MINI-WRAS) and the difference of the detector output signals was obtained and is shown in Fig. 8. The correlation was strong, Pearson’s correlation coefficient r was 0.904. The best-fit curve was a second-degree curve (with coefficient of determination R2 = 0.838) and was chosen because of its better representation of both very low and very high number concentrations, compared to the line, with its coefficient of determination R2 = 0.817. The quadratic model was better at determining the number concentrations of particles, although the 95% confidence intervals of the function parameters were wider, compared to the linear model. Therefore, this model was used in the main microcontroller of the ionisation nanoparticle detector to estimate the number concentration of particles, for the purpose of future measurements. Quadratic and linear model coefficients, standard errors and their 95% confidence intervals (CI) are shown in Table 5.

Fig. 8
figure 8

The correlation between the number concentration of NPs on the difference of the ionisation nanoparticle detector output signals — model comparison

Full size image
Table 5 Correlation equations — model comparison
Full size table

To improve the sensitivity of the device, the user is able to define the averaging time before the measurements. However, increasing the averaging time extends the time of a single measurement, i.e. reduces the frequency of the results obtained and displayed on the detector’s screen. An optimal solution for a specific user must be found in this case. To influence the sensitivity of the ionisation nanoparticle detector, the reference voltage in the electronic system can be reduced using external reference voltages.

Conclusions

Taking into account all the measurements carried out and the results, it can be concluded, that the ionisation nanoparticle detector, whose principle of operation and construction is based on a modified smoke detector, is a device capable of detecting NPs in the air and approximates their concentration (after appropriate calibration) in the range of 10,000–400,000 particles/cm3. The correlation obtained for measurements done in the first stage (without aerosol neutralisation) represents a good fit (r > 0.85). A strong correlation (r > 0.90) was also observed in the case of measurements done in the second stage (with neutralised aerosol). The ionisation nanoparticle detector follows the trend but does not always reach the exact maximum values. The process of neutralising particles can affect the way they interfere with ions in the active chamber of the detector. The results presented in the current article show that neutralised solid particles, after entering the active chamber, attract ions present in the chamber to their surface to a lesser extent. This results in a smaller change in the electrostatic potential measured at the floating electrode. As a result, the detector detects particles smaller than 100 nm, representing a smaller percentage of all particles. Taking all this into account, the ionisation nanoparticle detector meets the requirements of a low-cost instrument for the measurement of NPs’ number concentration. It can be used by companies where nanomaterials are used in production processes (production of composite materials and pharmaceuticals) or where there is a risk of their uncontrolled formation as a result of failure (high-temperature processes, crushing and grinding processes, paint coatings, etc.) and laboratories where nano-objects are used (biotechnology, surface functionalisation, etc.). However, a calibration to the specific process could be required, due to the differences in the measurements for neutralised and non-neutralised aerosol.