1. Introduction
In the European Union (EU), solid particle number (SPN) emissions are regulated since 2011 for light-duty vehicles (passenger cars and light commercial vehicles), when the Euro 5b emission standard came into force [
1,
2]. At first, the SPN limit was applicable only to vehicles applying compression ignition engines, but in 2014 (with Euro 6) it was extended to all light duty vehicles applying a direct into the cylinder fuel injection [
3], thus covering also positive ignition gasoline direct injection (GDI) vehicles. Since then, many countries in Asia have also introduced the SPN limit to all vehicles or vehicles with direct injection engines only [
4]. Currently the SPN limit is 6 × 10
11 #/km.
The SPN measurement methodology was developed by the United Nations Economic Commission for Europe (UNECE) coordinated Particle Measurement Programme (PMP) informal working group. The original aim was to have a good repeatability and reproducibility at low emission levels typically seen downstream of diesel particulate filters (DPFs) [
5]. The so called “PMP measurement system”, which is prescribed in the regulations, comprises a volatile particle remover (VPR) and a particle number counter (PNC). In the VPR, there is a hot particle number diluter (temperature 150 to 400 °C), a non-catalyzed evaporation tube (temperature 350 °C) and another cold diluter. The PNC has around 50% counting efficiency at 23 nm in order to include the primary soot particles (with particle sizes around 20 nm) as well as to avoid counting of volatile particles (volatile artefacts) formed between the VPR and PNC often detected in the sub-23 nm range [
6,
7,
8,
9].
Nevertheless, there has lately been an increasing interest in the SPN emissions in the particle size range below 23 nm. Notable levels of <23 nm diameter solid particles were reported for various vehicle/engine applications with various measurement methods [
9,
10,
11,
12,
13]. The excess of SPN emissions measured in the 10–23 nm range compared to the >23 nm emissions, can reach levels of over 500% for certain engine and vehicle applications [
14,
15]. Usually, the excess sub-23 nm particles for GDI and DPF equipped Diesel vehicles varies between 0% and 50%, and for port fuel injection (PFI) vehicles up to 140% [
9,
13,
16,
17]. In light of these findings, the decrease of the lower diameter of 23 nm of the current regulations to 10 nm for future regulations is of interest [
4]. The new methodology needs to be economical and, additionally, reproducible for <23 nm measurements [
18].
The discussion for the new methodology started in 2013 at the 28th PMP group [
19]. The discussion focused on keeping the link to the SPN
23 methodology, keeping existing systems as similar as possible, having low particle losses at the sub-23 nm range but at the same time avoiding volatile artifacts [
17]. The main differences to the current SPN
23 methodology, which was briefly described above, are the requirement of a catalyzed evaporation tube in the VPR and a 65% counting efficiency at 10 nm for the PNC [
20]. The catalyzed evaporation tube is required in the system to avoid volatile artefacts [
21,
22,
23]. The new methodology which counts solid particles from approximately 10 nm, hereafter referred to as SPN
10 methodology, was approved during the 81st session of the UNECE Working Party on Pollution and Energy (GRPE), in June 2020, as an amendment to the global technical regulation (GTR 15) [
24]. Details can be found in “2. Materials and Methods” section or in the literature [
4].
The repeatability and reproducibility of a method is important when results from different laboratories are compared and when emission limits are set [
5,
25]. Since the first comparison exercise in 2007 [
5], the current SPN
23 methodology has been extensively evaluated [
26,
27]. A few studies have applied the new methodology and others have compared instruments complying with the new methodology [
28,
29]. However, there are no inter-laboratory correlation studies about assessing the reproducibility of the recently introduced methodology. In order to assess the reproducibility of the new SPN
10 methodology and compare it with the current SPN
23, an inter-laboratory comparison exercise in nine different emission laboratories was conducted. One vehicle with reference equipment measuring from 23 nm and 10 nm was circulated in 7 European and 2 Asian laboratories. Cold and hot start type approval cycles were run at each laboratory. This paper presents the basic findings.
2. Materials and Methods
2.1. Overview
Laboratories in Europe and Asia measured the emissions of the same “reference” vehicle, using their own systems complying with the current 23 nm regulation and also using the same “reference” systems that were sent to them (including both 23 and 10 nm counters).
The inter-laboratory comparison exercise (ILCE) was done in two parts: the first part (Part I) included seven European emission laboratories (Lab 1 to Lab 7) and the second part (Part II) included one European laboratory and two Asian laboratories (Lab A to Lab C). The JRC participated in both parts conducting the measurements at the beginning and at the end of each part. The two parts are not comparable because maintenance of the car, including oil change, took place, plus different fuels were used at the two parts. Details will follow. The results are anonymized; the participants in alphabetical order were: BOSMAL (Bielsko-Biala, Poland) [
30], Empa (Dübendorf, Switzerland), Eni (S. Donato M.se, Italy), Ford (Aachen, Germany), JRC (Ispra, Italy), PSA (Stellantis) (Carrières sous Poissy, France), Tokyo Denki University (Tokyo, Japan), VETC (Xiamen, China), and VTT (Espoo, Finland).
The companies that provided the instruments (systems) for the measurement setup in alphabetical order were: Airmodus Oy (Helsinki, Finland), AVL List GmbH (Graz, Austria), Horiba Europe (Oberursel, Germany), and TSI GmbH (Aachen, Germany). In addition, JRC provided the remaining necessary parts of the sampling setup (e.g., tubing, splitters, pumps).
2.2. Vehicle
The vehicle was a 2017 Opel Astra turbo with a gasoline direct injection (GDI) engine (displacement 999 cm3, rated power 77 kW). It had manual transmission, front wheel drive and a three-way catalyst as exhaust aftertreatment (no particulate filter) and it complied with the Euro 6b emission standard (i.e., fulfilled the limit of 6 × 1012 #/km). The curb weight was 1213 kg (without the driver), the maximum payload was 637 kg, and the test mass was set to 1451 kg.
2.3. Fuel
In the first part (Lab 1 to Lab 7), E5 market fuel from the same batch was delivered to the laboratories, while in the second part (Lab A to Lab C) each laboratory acquired their own European reference type E10 fuel, see fuel characteristics data in
Table 1.
2.4. Measurement Setup
The measurement setup is given in
Figure 1. All systems were sampling from the dilution tunnel with constant volume sampling (CVS). Within the grey dashed lines, in addition to the vehicle, are the instruments that were shipped to the laboratories.
Each laboratory used their own 23 nm system, compliant with the current regulation [
24]. Their VPR included a non-catalyzed evaporation tube (ET) and a PNC with a counting efficiency of 50% (±12%) at 23 nm and over 90% at 41 nm (PNC
23). Externally, a PNC having 65% (±15%) counting efficiency at 10 nm and >90% at 15 nm (PNC
10) was connected. This PNC
10 was the model A20 from Airmodus Oy (Helsinki, Finland). In parallel, a 3792E from TSI GmbH (Aachen, Germany) with the same specifications was connected as a monitor/control counter. Due to some issues in one of the reference system, the second 3792E was used very often at that system. Thus, the data set recorded with 3792E in parallel with the A20 was limited, and the results are not presented. Nevertheless, when the two PNCs were measured in parallel, their differences were within 3% (±2%). The whole system is abbreviated as ET, and depending on the lower size ET
23 or ET
10.
Two systems measuring both 23 nm and 10 nm PNCs were provided to the laboratories and were measuring in parallel from the dilution tunnel. Their VPR had a catalyzed evaporation tube (CS). These systems are named hereafter CSA and CSB. CSA, the AVL Particle Counter (APC) model: 489, was provided by AVL (Graz, Austria) and included two PNCs from AVL, while CSB, SPCS 2300, was provided by Horiba Europe (Oberursel, Germany) with a PNC10 from TSI. PNC23, provided by TSI (model 3791), was connected to CSB. All systems were returned to the instrument manufacturers for maintenance between the two parts of the ILCE.
The setup, which is based on the light-duty vehicle emission regulations, allows a direct comparison of the instruments which were measuring simultaneously. However, comparison of the results between different laboratories is influenced by the stability of the vehicle and the particle losses (thermophoretic, agglomeration) in the tubing between the vehicle and the tunnel. This tubing is not standardized and it can be up to 6-m long heated or insulated, with different quality of the internal surfaces.
2.5. Test Cycles
The test protocol included three cold start and five hot start world harmonized light vehicles test cycles (WLTCs). The cycle consists of four phases (low, medium, high, extra high, abbreviated as WLTC 1–4 respectively). At the beginning of the cold start tests the oil temperature was within 2 °C of the ambient temperature (23 °C), while at the hot start tests the oil temperature was >80 °C.
2.6. Calculations
All data were delivered to JRC and anonymized. At a first step, the data in each laboratory were screened manually for measurement errors, such as incomplete time series or high background particle concentration levels. In Lab 2, the dilution for the external PNC10 of the ET system was determined erroneously and the reported emissions were extremely high. Lab 5 had very high background concentrations. In Lab 1, due to communication issues, the CSA,23 had often zero time-series in the reported concentrations. These results were excluded from the further analysis as they were clear errors, which would have been detected and excluded in normal emission measurements.
In CSB, an internal filter was plugged gradually during the measurement campaign and the results with and without plugged filter deviated as much as 20%, as it was determined by the instrument manufacturer during the intermediate validation of the system. The plugging happened twice. In Lab 3, a leak was detected in the CSB internal PNC10 and it was replaced with another PNC10 (TSI 3792E, the monitor PNC at the ET system), which was within 5% of the PNC10 connected to the laboratories won systems. These problems were considered as part of the experimental uncertainties and the results were included in the final analysis. This means that the results of CSB will give the high range of expected reproducibility, as they include a 20% internal variability, plus a 5% variability of the two PNC10 that were used.
In the second step, a statistical analysis of outliers was conducted to further screen the data validity. Cochran and Grubbs tests were applied to test the emission results of all the systems for variance and average outliers, respectively [
31]. Although, CS
B,23 cold start WLTCs from Lab 7 were determined to be an outlier in the Cochran test, the results were kept because they were considered to represent a possible variability, as CS
B was sent to Lab 7 directly after maintenance.
The particle number emissions for a system (ET, CS
A, CS
B) for a specific lower size (23 nm, 10 nm), SPN
size, (#/km), were calculated with the following equation:
where V
CVS (m
3) is the total volume of the dilution tunnel and D (km) the cycle distance. PCRF (-) is the average particle number concentration reduction factor at 30 nm, 50 nm, and 100 nm for the specific system. PNC
size (#/m
3) is the particle number counter concentration in a specific system with a lower size of 23 nm or 10 nm. Both PNC
size and V
CVS were normalized (corrected) to the same temperature and pressure conditions.
The excess sub-23 nm (%) particles for each system was calculated as:
The reproducibility coefficient of variance was calculated following ISO 5725-2 (1994). Details can be found elsewhere [
31]. The JRC was measured at the beginning and end of each part (four measurement campaigns). The data were treated similarly to the other laboratories and only one measurement campaign was included in the data-analysis for each ILCE part (the end measurements of each part).
To assess the stability of the vehicle emissions, the results of JRC were plotted in chronological order, but normalized to (i.e., referred to) the first tests of the first ILCE part. To assess the comparability of the systems, their differences from their mean were calculated at each laboratory. Finally, to assess the inter-laboratories variability (or between laboratories variability) the differences of each system from the mean of all laboratories were calculated.
3. Results
3.1. Vehicle Stability
The vehicle emission stability was monitored by measuring the SPN emissions at the JRC at two different vehicle emission laboratories (VELA 2 at the start and end of the first part of ILCE, VELA 1 at the start and end of second part). As it was mentioned in 2. Materials and Methods, maintenance, including oil change, was conducted after the European part, and different fuel was used between the two parts (E5 part I, E10 part II). Thus, the emissions of the two campaigns are not comparable. Nevertheless, the emissions of each system and size are normalized to the beginning of the ILCE (Part I: Start) in
Figure 2.
There is a decreasing trend in the SPN with a difference reaching 10–30% between start and end of ILCE. In the European part, the start and end measurements had a difference of <10%, suggesting relatively stable vehicle functioning over the European ILCE part. In the Asian part (Part II), the start emissions were quite similar with the end of the European part (differences < 10%). However, at the end another 5–15% decrease was noted. The lower percentages are based on the CSA system that was the same system during the whole ILCE. The higher percentages with the ET system are probably due to the additional uncertainty of the different facilities (and systems) that were used between the two parts of the ILCE. Concluding, based on CSA, the differences between start of Part I and end of Part II were 10–20%, while the differences of the averages of each part were 6–10%.
3.2. WLTC Absolute Emissions
Figure 3 plots the results for each laboratory and system for the cold start WLTCs. The error bars show the intra-laboratory standard deviation (repeatability) for each system. The cold start WLTC emission levels, based on the reference ET
23 systems, ranged from 3.3 × 10
12 to 5.6 × 10
12 #/km at the European part (Part I) and 2.5 × 10
12 to 2.8 × 10
12 #/km at the Asian part (Part II). The ET
10 results were 3.8–5.6 × 10
12 #/km at Part I and 2.6–3.2 × 10
12 #/km at Part II. The hot start WLTC emissions were approximately half of the cold start emissions.
The repeatability of the measurements, shown as error bars, was on average 5% for the 23 nm measurements with all three systems (reaching 11% at Lab 3), and 4% for the 10 nm measurements (reaching 8% at Lab 7).
Appendix A gives the normalized differences for each measurement system from the mean at all laboratories, as an indication of the inter-laboratories variability with each system (or between-laboratories variability). They were within ±30%. Practically this means that the reproducibility (combination of repeatability and inter-laboratories variability) is determined by the inter-laboratories variability.
3.3. Sub-23 nm Particles
Figure 4 shows the excess of sub-23 nm particles determined by all systems for the cold start WLTCs.
The excess of sub-23 nm particles with the ET systems were in general lower (8 to 19% percentage points) than those determined with the reference systems CSA and CSB. One reason is the non-optimized sampling position of the PNC10 that could have resulted in high sub-23 nm particle losses. PNC10 was connected externally to the ET systems, a position that in some cases included valves, flowmeter, and long tubing. Another reason could be calibration uncertainties of the PNC23 counters, or even underestimation of the PNC10. Finally, the ET systems had evaporation tube instead of catalytic stripper (as the reference systems). It is possible that downstream of the evaporation tube some material condensed on the particles resulting in higher mean sizes and, consequently, lower sub-23 nm ratio.
What is of particular interest here is the variability of the sub-23 nm values between laboratories. Based on the CS
A system, in the first part of the ILCE, the excess of sub-23 nm particles varied from 13% to 23% for the cold start WLTC (
Figure 4a), and 32% to 46% for the hot start WLTC (
Figure 4b). In the second part of the ILCE, the excess of sub-23 nm particles varied from 27% to 32% for the cold start WLTC (
Figure 4a), and 40% to 56% for the hot start WLTC (
Figure 4b). The average sub-23 nm value was higher for the second part of the ILCE than for the first part for all systems. Based on CS
A, the fraction was 10% points higher in the second part. This suggests that not only there was a decrease in the emission levels (discussed in
Figure 3), but also in the particle size from the first to second part of the ILCE. The higher fractions at both cold and hot start tests were always from the same laboratories indicating that the facilities setup might have played a role as well (e.g., tubing length from vehicle to tailpipe, heating).
Based on the values of
Figure 4, the standard deviation of sub-23 nm values of the different laboratories was around 10–20% for ETs, 4–8% for CS
A, and 8–15% for CS
B. When considering the problems detected with CS
B and the variability between the different ET systems, the accuracy of the determination of the sub-23 nm values may be assumed from the CS
A results (4% to 8%).
3.4. Comparisons of Measurement Systems
Figure 5 plots the differences of each system from the mean of the three systems for each laboratory. Each color refers to a different system. The calculations were done separately for the 23 nm systems (dark colored bars) and the 10 nm systems (light colored bars). The error bars show min-max values from the cold start and hot start WLTCs.
In general, the differences of the ET and CS
A systems were comparable (i.e., within ±10% in most cases) considering the variability between laboratories (around ±30%) (see
Figure A1 and
Figure A2 in the
Appendix A). CS
B had slightly higher variability (up to ±15%) due to internal problems during the ILCE. What is important though is that the variability of the 10 nm systems is not higher than the 23 nm systems variability (compare light vs. dark colored bars).
3.5. Reproducibilities
Table 2 gives the overall results with each system and the relevant reproducibility values. The mean emissions were clearly lower in the second part of the ILCE. For example, for the cold start WLTC the average emissions were 28% to 39% lower; the exact difference depending on the system. For the hot start WLTCs the emissions were 14% to 28% lower. The differences were higher, but in line, with the stability tests conducted at JRC (see
Figure 2), where a 10–30% decrease was noticed. As it was mentioned in 2. Materials and Methods, the car was maintained, and the oil was changed between the two parts. In addition, different fuel was used at the two parts.
Focusing on the details of
Table 2: The reproducibility of the current methodology measuring >23 nm using evaporation tube was 19–24% in the first part, but 7–19% in the second part of the ILCE. For the 10 nm counters the values were at similar or even lower levels (17–18% and 10–11% respectively). This was expected as the 10 nm emissions were measured with the same PNC
10 in each lab while the 23 nm emissions with different PNC
23 in each lab. The best performing system CS
A had much lower reproducibility values (11–14% at the first part and 14–25% at the second, for both 23 nm and 10 nm measurements). The second system was in general in the 15–36% range. CS
B variability was affected by the internal filter problems. Although the reproducibility results of the two parts were in line, the numbers of the first part are more representative as there were more laboratories involved with more tests: 65 test (first part) vs. 23 tests (second part).
3.6. Emissions in WLTC Phases
Figure 6a plots the average SPN emissions for each phase of the cold start WLTC with each system (ET, CS
A, CS
B) for each part of the ILCE. Part I (European) with dark colors, Part II (Asian) with light colors. The emissions were more than two times higher in the first phase of the cycle compared to the other phases. At the hot start tests the emissions were relatively equal between the phases (no figure shown). The previously discussed (
Figure 3) differences between the two parts of the ILCE can also be seen here. Here it is evident that the biggest differences happened on the first phase of the WLTC with the differences gradually decreasing. At the fourth phase of the cycle the differences between the ILCE parts were within experimental uncertainty. The error bars were higher at the cold start phase (20–30%) becoming smaller as the engine was reaching stable (hot) operating conditions (2–20%).
Figure 6b plots the excess sub-23 nm particles in the same format as
Figure 6a. There is an increasing trend of the sub-23 nm percentages along the cold start WLTC phases: the sub-23 nm value is lowest in the first phase (around 5%) and increases toward the fourth phase (around 40%). The percentages were slightly higher in the second part of the ILCE compared to the first part.
These results show that the higher variability at the cold start phase comes from the different absolute levels, and not from the size of particles, which was well above the cut-off size of both systems. A small size close to the cut-off curves of the systems could result in high variability.
4. Discussion
The main objective of this study was to assess the repeatability and reproducibility of the new 10 nm methodology and compare them with the current 23 nm methodology. The selection of the vehicle for this purpose was important. As the new methodology was aiming at GDI vehicles (and in the future other spark ignition vehicles), it was necessary to select a GDI vehicle. Previous inter-laboratory studies with Diesel vehicles equipped with DPF were affected by the variability of the vehicle emissions due to the DPF fill state [
5,
26]. Although the gasoline particulate filters (GPFs) are not so sensitive to the fill state due to the continuous passive regeneration, the experience on the long term behavior of GPFs was limited at the time of the ILCE design. Yet, a reduction of emissions over time due to ash accumulation in the filter was reported in some studies [
32], and thus, it was decided to select a GDI without GPF to avoid these issues.
The emission levels of the selected GDI vehicle (without GPF) covered a range of 1 to 15 × 10
12 #/km (the lower values for hot start tests). Such emission levels were typical for Euro 6b GDI vehicles [
3,
33,
34,
35,
36,
37]. Lower values were measured at the second part of the ILCE after maintenance, oil change, and use of E10 instead of E5. The stability tests of JRC with the same system showed a 10–30% difference. It is possible that the lower emissions reflected the higher oxygen content of the E10 fuel at the second part. A moderate increase in fuel oxygen content in low oxygen content fuels is connected to a decrease in emissions for GDIs [
38,
39,
40]. The lack of adequate lubricant conditioning (driving > 500 km) may also explain the decreasing trend in the second part [
41]. Even though the two parts of the ILCE were treated separately the conclusions would be the same treating them all together. The only difference would be that the reproducibility values would be higher, but any relative comparisons (e.g., between systems, 23 nm vs. 10 nm) would be similar. For example, the 10 nm CS
A system had means of 5.0 and 3.5 × 10
11 #/km at Part I and Part II respectively for the cold WLTC, with similar reproducibility values of 14% at both parts. Combining both parts, the mean would be 4.5 × 10
11 #/km with 21% reproducibility. The 23 nm CS
A system had means of 4.2 and 2.7 × 10
11 #/km at Part I and Part II respectively, with reproducibility values of 14% at both parts. Combining them, the mean would be 3.7 × 10
11 #/km with 23% reproducibility. The 10 nm ET system reproducibility would be 27% (from 10–17%).
The sub-23 nm particles compared to the >23 nm emissions were 10–50%. The lower values were noted during cold start, which means larger particles compared to the hot engine operation. In the second ILCE part, the sub-23 nm to >23 nm ratio was higher, indicating a lower mean particle size decrease probably due to the higher fuel oxygen content. Moreover, the oil change might have affected the emission characteristics [
42]. In any case the sub-23 nm values were in agreement with the literature [
3,
33].
In addition to the laboratories own equipment, two reference systems were also circulated. The difference of the three systems from their means were within ±10% in almost all cases. Higher differences were noticed with one system due to internal issues during the ILCE (±15%). Such differences are well within the expected differences of measurement systems. In the literature, differences of 20% have been reported for both 23 nm [
2,
43,
44] and 10 nm systems [
29,
45,
46].
One of the most important results of this study is the reproducibility values that were calculated for the different systems. The values were 10–25%, with the exception of the CS
B system that reached 35% due to the above mentioned internal issues. The maximum reproducibility coefficient of variance in this study was at the same level or slightly lower compared to studies measuring from the CVS, which reported values of 35–45% [
5,
27]. In both of the cited inter-laboratory studies, the diesel particle filter condition affected the repeatability. In a recent study with a diesel vehicle emitting close to the limit, the reproducibility was 22% [
25]. Another study with a gasoline vehicle reported 29% reproducibility [
47]. The following factors may have affected the reproducibility [
45]: at emission levels >10
12 #/km, agglomeration in the line between the vehicle and the dilution tunnel can result variability of SPN
10 between different laboratories [
45]. Similarly, thermophoretic losses are important and can be different depending on the transfer line temperature. In the ILCE the labs applied different setups: from mixing tee within one meter from the tailpipe up to 9 m tube. Some parts of the tubes were insulated and some were heated at various temperatures (from 50 up to 170 °C). Finally, the stability of the vehicle itself plays an important role. Treating the two parts separately, it was found that the stability was around 10% in the first part and 10–15% in the second part. The vehicle variability is included in the reproducibility values.
Our results proved that the 10 nm methodology has similar or even lower reproducibility than the current 23 nm methodology. In fact, this is not surprising. The 10 nm counters are less sensitive to the surface properties and chemical composition of particles [
48,
49], compared to the 23 nm counters [
50,
51]. In addition, the catalytic strippers remove the condensing species, minimizing the impact of condensation growth on the counting efficiency of the counters [
23]. Nevertheless, the concentration of particles below 23 nm was practically equal or lower than the concentration of particles above 23 nm for the GDI vehicle of our study. This indicates mean particle sizes around 20 nm or higher i.e., above the 10 nm size of the new methodology. Size distributions peaking at 10 nm might result in higher variability because the counting efficiency of the systems will be different at small sizes. For example, for size distributions with mean sizes >40 nm the differences between systems are expected to be within ±10%, but below 20 nm the differences can be >30%
5. Conclusions
Seven European and two Asian laboratories measured the solid particle number emissions of a reference gasoline direct injection (GDI) vehicle with their own equipment and two reference systems that circulated to all laboratories. All systems were equipped with both 23 nm and 10 nm counters. The laboratory systems had an evaporation tube for thermal pre-treatment while the reference systems applied catalytic strippers. The aim of the inter-laboratory exercise (ILCE) was to compare the reproducibility of the new 10 nm methodology to the current 23 nm methodology.
The European (Part I: 7 European laboratories) and Asian (Part II: 2 Asian laboratories and one European) parts of the ILCE were treated separately because maintenance (including oil change) of the vehicle took place between the two parts. Furthermore, E5 market fuel from the same batch was delivered to the European laboratories, while the Asian laboratories acquired their own reference type E10 fuel. The solid particle number (SPN) emissions were 20–35% lower at the Asian part II than at the European part I. The highest differences between the ILCE parts were detected at the first phase of the cold start cycle, while the differences at the last phase were minimum (hot engine).
In general, the intra-laboratory variability for the 23 nm SPN systems were within ±8%, while for the 10 nm systems it was within ±10%. On the other hand, the inter-laboratories reproducibility was much larger up to 25% for both 23 nm and 10 nm measurements. Thus, the 10 nm methodology uncertainty was similar with the 23 nm methodology uncertainty. More specifically, the reproducibility was 7–24% for the current 23 nm methodology with evaporation tube. The addition of a 10 nm counter in the measurement system did not increase the levels (10–18%). The first reference system, CSA, had the lowest variability with values ranging between 13–25% (23 nm) and 11–20% (10 nm). The variability of the second system, CSB, was 17–36% (23 nm) and 15–28% (10 nm) due to some technical issues during the ILCE. The higher variability of all systems were typically detected at the second (Asian) part due to the smaller number of laboratories (3 laboratories), compared to the first part (7 laboratories). Within the test cycles, the variability was highest in the first phase of the WLTC, decreasing toward the last phase of the cycle, for both cold and hot start WLTC. This trend was more distinct in the first part of the ILCE where the absolute cold start levels were higher than in the second part.
The excess of particles below 23 nm were 18% (±5%) in the European part and 30% (±3%) in the Asian part for the cold start WLTC based on the first reference system CSA. The values for the hot start WLTC were 39% (±7%) and 49% (±8%) respectively. The difference between the two parts comes from the different mean particle sizes. The differences of sub-23 nm percentages between cold and hot start cycles were due to the low sub-23 nm values (i.e., large particles) during the first phase of the cold start WLTC, compared to hot start WLTC. The laboratory systems measured lower sub-23 nm fractions compared to the reference systems due to additional losses for the 10 nm counters in some laboratory systems, calibration issues, and probably due to condensational growth of the particles between the evaporation tube and the counter.
The main message of this study is that, according to the presented data, the change of the lower detection limit of systems from 23 nm to 10 nm, does not deteriorate the reproducibility of the solid particle number measurement.