Large daytime signals of N 2 O 5 and NO 3 inferred at 62 amu in a TD-CIMS : chemical interference or a real atmospheric phenomenon ?

Dinitrogen pentoxide (N2O5) and the nitrate radical (NO3) play important roles in atmospheric chemistry, yet accurate measurements of their concentrations remain challenging. A thermal dissociation chemical ionization mass spectrometer (TD-CIMS) was deployed to an urban site in Hong Kong to measure the sum of N 2O5 and NO3 in autumn 2010 based on the signals of NO −3 at 62 amu which has also been adopted in previous studies reported in literature. To our surprise, very large signals of N 2O5 + NO3 were frequently observed at 62 amu in the daytime, with equivalent N2O5 + NO3 mixing ratios in the range of 200–1000 pptv. To investigate this unusual phenomenon, various interference tests and measurements with different instrument configuration were conducted. It was found that peroxy acetyl nitrate (PAN) contributed to measurable signals at 62 amu, and more importantly, this interference increased significantly with coexistence of NO2. Nitric acid (HNO3), on the other hand, had little interference to the detection of N 2O5/NO3 via the NO−3 ion in our TD-CIMS. According to the test results, the interference from PAN and NO 2 could have contributed to 30–50 % of the average daytime (12:00–16:00, local time) N2O5 + NO3 signal at our site. On the other hand, evidence exists for the presence of elevated daytime N 2O5, in addition to the daytime signal at 62 amu. This includes (1) daytime N2O5 measured via the I(N 2O5) cluster ion with an unheated inlet, which was subjected to minimum interferences, and (2) observation of elevated daytime ClNO 2 (a product of N2O5 hydrolysis) during a follow-up study. In view of the difficulty in accurately quantifying the contribution from the interferences of PAN and NO 2 and untested potential interfering chemicals in the real atmosphere, we caution the use of 62 amu in the TD-CIMS for measuring ambient N 2O5 in a high NOx environment like Hong Kong. Additional studies are needed to re-examine the daytime issue using other measurement techniques.


Introduction
The nitrate radical (NO 3 ) and dinitrogen pentoxide (N 2 O 5 ) play important roles in the nocturnal tropospheric chemistry.NO 3 is among the most important oxidants in the atmosphere, particularly for biogenic hydrocarbons and sulfurcontaining compounds (Atkinson, 1991).N 2 O 5 has long been recognized as a key intermediate in the transformation of nitrogen oxides (NO x = NO + NO 2 ) to aerosol nitrates (Riemer et al., 2003;Aldener et al., 2006;Chang et al., 2011;Brown and Stutz, 2012).Recent studies have also demonstrated an important role of N 2 O 5 hydrolysis at night in chlorine activation and the subsequent effect on the next-day's ozone formation (Osthoff et al., 2008;Simon et al., 2009;Thornton et al., 2010).
Due to the low ambient abundances and high reactivity, accurate measurements of atmospheric N 2 O 5 and NO 3 have been challenging.Based on the strong absorption of NO 3 in the visible spectrum at 662 nm, several optical techniques have been developed to measure the ambient NO 3 , including long-path differential optical absorption spectroscopy (DOAS) (Platt et al., 1980;Atkinson et al., 1986), cavity ring-down spectroscopy (CRDS) (Brown et al., 2001(Brown et al., , 2002)), laser-induced fluorescence (LIF) (Wood et al., 2003(Wood et al., , 2005;;Matsumoto et al., 2005), and cavity enhanced absorption spectroscopy (CEAS) (Venables et al., 2006;Langridge et al., 2008).N 2 O 5 is determined by using a heated channel to decompose it into NO 3 or from the calculation according to the fast equilibrium between N 2 O 5 with NO 3 and NO 2 .
Another technique for detecting ambient N 2 O 5 and NO 3 is chemical ionization mass spectrometry (CIMS) which combines the ion-molecule chemistry with mass spectrometry detection.This technique was originally used in the laboratory to study the heterogeneous uptake kinetics of N 2 O 5 and ClNO 2 (e.g., Hu and Abbatt, 1997;Thornton et al., 2003;Thornton and Abbatt, 2005), and later on was applied in field measurements (Slusher et al., 2004;Zheng et al., 2008).The fundamental of this method is the reaction of I − (the reagent ion) with N 2 O 5 (and/or NO 3 ) forming the NO − 3 ion that can be detected at 62 amu.Previous laboratory studies suggested that the N 2 O 5 /NO 3 measurement at 62 amu may be subject to interferences from other N-containing trace gases such as HNO 3 , HO 2 NO 2 , ClONO 2 and others (see references listed in Table 1).On the other hand, a field inter-comparison of a TD-CIMS and a cavity ring-down system showed a high degree of correlation between the two methods (Huey, 2007;Chang et al., 2011) indicating the capability of this method in measuring N 2 O 5 in the atmosphere.Although non-negligible and varying background signals at 62 amu have been observed during field studies (Chang et al., 2011), there have been no reports of detailed assessment of these potential interferences under different atmospheric conditions and for different configurations of CIMS.Kercher et al. (2009) developed a method to detect N 2 O 5 via the I(N 2 O 5 ) − cluster ion at 235 amu with an unheated inlet to address the interference at 62 amu in their CIMS, however this method has its own limitations including a lower sensitivity and larger impact of water vapor than that at 62 amu.
In autumn 2010, a TD-CIMS (thermal dissociation-CIMS), which is the same type used by Slusher et al. (2004) and Huey (2007), was deployed to an urban site in Hong Kong which is characterized by large quantities of NO x , ozone and particulate matters.Unexpectedly, large signals of N 2 O 5 + NO 3 inferred at 62 amu were frequently observed during the daytime in our study.To investigate this unusual observation, we have conducted a series of laboratory and field tests, including testing interferences individually and in combination from PAN, NO 2 , O 3 , HNO 3 , detection of N 2 O 5 by using a cold inlet via detection of the I(N 2 O 5 ) − cluster ion at 235 amu, and examination of daytime ClNO 2 which is a product of N 2 O 5 hydrolysis.In this paper, we will first present the ambient observations, and then the detailed test results on interferences and other evidence for daytime N 2 O 5 .A surprising result from the tests is that PAN+NO 2 can contribute a large interference to the TD-CIMS signal at 62 amu, which has not been reported in previous studies.Despite large chemical interference, we show that the observed  daytime N 2 O 5 signal may be in part due to a real contribution from NO 3 and N 2 O 5 .

Measurement site
The field study was conducted in the urban center of Hong Kong (22 • 18 N, 114 • 11 E; ∼ 15 m above sea level).The sampling site was located on the rooftop of a seven-story teaching building (about 20 m above the ground) on the campus of the Hong Kong Polytechnic University (PolyU).To the southeast of the campus, there is a cross-harbor tunnel in the north-south direction with a large flow of vehicles, especially during rush hours (see Fig. 1).Victoria Harbor is located about 1 km south of the measurement site.Thus emissions from vehicles and marine vessels are the most important local anthropogenic sources.The sampling site is surrounded by business districts, tourism and residential areas, with no large industrial sources nearby.The field measurements were carried out from 15 October to 4 December 2010, which is the season with the most severe photochemical pollution in Hong Kong (Wang et al., 2009).

The CIMS apparatus
A TD-CIMS was deployed to measure the sum of N 2 O 5 and NO 3 in this study.The system was developed at the Georgia Institute of Technology and is based on a soft and selective ionization process resulting from the reaction between a reagent ion and the target compounds, with the generated ions detected by a mass spectrometer.In the present study, the measurement method and operating parameters of TD-CIMS were the same as those described by Slusher et al. (2004) which was configured to simultaneously measure ambient N 2 O 5 and PANs.The schematic diagram of our TD-CIMS is shown in Fig. 2. I − , which was produced from passing a flow of 2 sccm of 0.3 % CH 3 I/N 2 through an alpha ion source (Po-210), served as the reagent ion.Ambient air samples were drawn through a PFA-Teflon tube (I.D., 9.5 mm; O.D., 12.7 mm; length, 2 m) to the CIMS at a flow rate of 1.5 standard liters per minute (SLPM).To reduce the residence  et al., 2004), and under these conditions more than 99 % of the N 2 O 5 would decompose into NO 3 with an ambient NO 2 level of 40 ppbv.The produced and the original NO 3 then reacted with I − to produce NO − 3 ions in the flow tube, which were subsequently detected and quantified by a quadruple mass spectrometer (Extrel 150-QC, with a FWHM resolution of 2000) at 62 amu with a temporal resolution of 6 s.The NO − 3 signal is proportional to the concentration of total NO 3 , the I − signal, the effective reaction rate constant between them, and the effective reaction time (Huey, 2007).Figure 3 depicts mass spectra of ambient air in urban Hong Kong obtained both during the daytime and at night, which clearly shows the I − signal at 127 amu, NO − 3 signal at 62 amu, CH 3 C(O)O − (PAN) signal at 59 amu, etc.The mass peak at 62 amu was well separated from that at 59 amu.
In the present study, the TD-CIMS instrument was calibrated once a week using the on-line N 2 O 5 synthesis method (Bertram et al., 2009).The calibration source was generated from the reactions of NO 2 with O 3 and subsequently NO 3 with NO 2 .In a glass reaction chamber in a commercially available calibrator (Model 6100, Environics) 90 sccm of 2.5 ppmv NO 2 in nitrogen gas was mixed with 480 sccm of 2.1-4.6 ppmv O 3 which was generated by UV photolysis of O 2 in zero air (Model 111, Thermo Environmental Instruments (TEI)).The reactions happened during a time period of one minute, and then the output was diluted to 6 slpm by zero air.The concentrations of the prepared N 2 O 5 were determined via the change in NO 2 concentrations after adding ozone, and conversely verified by the change in ozone after adding NO 2 .Zero air that was free from moisture and aerosols served as the diluent so as to prevent the hydrolysis of N 2 O 5 during the calibrations (see Fig. 2).NO 2 was monitored with a chemiluminescence analyzer equipped with a photolytic converter which ensured measurements of the true NO 2 (Ryerson et al., 2000;Xu et al., 2013).When inputting 36.7 ppbv of NO 2 and 170.4 ppbv of O 3 , the generated N 2 O 5 was 1.08 ppbv.Note that there could be slight overestimation in the concentration of N 2 O 5 standard determined in our study, because a small fraction (< 10 %) of N 2 O 5 could be converted into gaseous nitric acid (Bertram et al., 2009).The loss of total nitrogen in the calibration system was examined with a chemiluminescence NO x analyzer equipped with a molybdenum converter by comparing the signal with  Based on the relative standard deviation of the sample signal, the precision of our TD-CIMS was 3 % for 1000 pptv N 2 O 5 .The sensitivity of N 2 O 5 during the campaign was 2.8 ± 0.2 (mean ± SD) Hz pptv −1 .The instrument background was automatically measured for 2 min once an hour by adding a small flow (5 mL min −1 ) of NO (1000 ppm) to the sample flow (diluted to 9.2 SLPM), titrating NO 3 and thus N 2 O 5 .The background signal of the NO − 3 ion during this field study was 71.7 ± 36.0 (mean ± SD) Hz. Figure 4 shows the signals of NO − 3 (at 62 amu) with hourly automatic background determinations on 30 November, when the equivalent maximum concentration of N 2 O 5 + NO 3 was inferred.Clearly, the signals exhibited relatively low background compared to ambient N 2 O 5 + NO 3 during both day and night.According to three times the standard deviation of the background signal, the typical detection limit of N 2 O 5 for 6 s average time was estimated to be 39 pptv for our TD-CIMS.

Other instruments
In addition to the TD-CIMS, a large number of other instruments were deployed concurrently.Here, we briefly describe those that were used to aid the presentation of the N 2 O 5 and NO 3 data.O 3 was measured by a commercial UV photometric analyzer (Model 49i, TEI, USA).NO and NO 2 were analyzed with a chemiluminescence instrument (Model 42i, TEI) equipped with a photolytic NO 2 converter (Air Quality Design, USA) (Xu et al., 2013).Solar radiation was measured using a LI-200 Pyranometer Sensor (LI-COR, USA).The ambient RH and temperature were monitored with a RH/temperature probe (Model 41382VC/VF, M.R. YOUNG, USA).During the field measurements, the minute-average data of trace gases and meteorological parameters were collected in real time by a data logger (Model 8816, Environmental Systems Corporation, USA).

Observation results
The time series of hourly NO − 3 signals at 62 amu (after subtracting the background by adding NO) and apparent N 2 O 5 + NO 3 mixing ratios (hereafter referred to as N 2 O 5 + NO * 3 ) from 15 October to 4 December 2010 is shown in Fig. 5. Similar to the measurement results obtained in other locations, several nighttime signal peaks of NO − 3 were noticed in our study (e.g., 31 October, 1 and 10 November).However, very large signals of N 2 O 5 + NO 3 inferred by NO − 3 ion at 62 amu were frequently observed during the daytime (e.g., 23 and 24 October, 19, 22 and 30 November and 2 December).During the 50 day measurement period (excluding 11 November due to a lack of data), the average hourly apparent N 2 O 5 + NO 3 concentration was 86.9 (±85.6)pptv with the maximum value of 1033 pptv recorded at 15:00 LT (local time) on 30 November.The mean daytime apparent N 2 O 5 + NO 3 mixing ratio was 102.5 pptv (06:00-17:59 LT), even higher than the value of 71.3 pptv at nighttime (18:00-05:59 LT).
To learn more about the atmospheric conditions associated with the elevated daytime N 2 O 5 and NO 3 signals at 62 amu, we present six cases during which the hourly apparent N 2 O 5 + NO 3 values exceeded 400 pptv.Figure 6 presents the 5 min data of N 2 O 5 + NO * 3 , O 3 , O x (O 3 + NO 2 ), NO, NO 2 , RH, and solar radiation for these episodes.The ratio of NO 3 / N 2 O 5 , calculated based on the temperature dependent equilibrium among N 2 O 5 , NO 3 and NO 2 , is also given.Inspection of the figure reveals that the daytime apparent N 2 O 5 + NO 3 signal peaks appeared when both ozone and NO 2 were in high levels together with low levels of NO.However, daytime N 2 O 5 + NO 3 concentrations calculated using both steady-state (Osthoff et al., 2006) and non-steadystate approaches (McLaren et al., 2010) were much lower (by a factor of 1-100) than observations.The daytime concentrations of apparent N 2 O 5 + NO 3 in our study are also much higher than those of previous studies which reported daytime N 2 O 5 and NO 3 concentrations of only a few to over 10 pptv (Geyer et al., 2003;Brown et al., 2005;Osthoff et al., 2006).

Chemical interferences
A major drawback of the TD-CIMS technique is that there could be potential interferences to the selected NO − 3 ion that can give rise to significant background noise at 62 amu (Chang et al., 2011).Previous laboratory studies have proposed many possible ion-molecule reactions yielding the NO − 3 ion, as summarized in Table 1.As stated earlier, the detailed results of these interferences in the real atmosphere have not been reported to date.To examine the influences of these potential interferences to our N 2 O 5 + NO 3 measurements, a series of tests were conducted in various atmospheric conditions and in the laboratory during and after the measurement campaign.
The background determination by adding excess NO provided a first examination of the interferences.Such determined background reflects the interferences from compounds that are unreactive to NO, including most potential ones other than PAN and HO 2 NO 2 .As stated above, the background was periodically measured once an hour throughout the campaign.A considerable background signal in our CIMS measurements at 62 amu was indeed seen during the present study, and they also exhibited a diurnal pattern with higher values in the late afternoon (see Fig. 4).This indicates that the N 2 O 5 + NO 3 measurements via the 62 amu channel in our TD-CIMS were subject to some interference during the present study.However, the background signals were much lower than the total signal in ambient air.For the six cases with daytime N 2 O 5 + NO 3 concentrations exceeding 400 pptv, the instrument background on average only accounted for 10 % of the ambient values.This background signal is due to most of the gases shown in Table 1 and has been accounted for (i.e., subtracted from the total signals at 62 amu) in our final data.The interference from PAN and HO 2 NO 2 could not be determined because they also reacted with NO during zeroing.Their interferences are addressed below.

Interference of PAN
PAN is the most possible compound to interfere the TD-CIMS measurements in this study, not only due to its relatively high ambient abundances but also because it can escape from background determination by adding NO.In the previous studies of using a different technique of CRDS, higher N 2 O 5 signals than the steady-state predictions were usually observed with elevated mixing ratios of PAN (Brown et al., 2005;Osthoff et al., 2006).The interference from PAN to the NO − 3 signals in our TD-CIMS was evaluated post field campaign by adding synthetic PAN to zero air and in the ambient air samples.The PAN was generated from a PAN calibrator (Meteorologie Consult GmbH), which is based on the reaction sequence of NO and acetone in ultrapure air with a Penray lamp.The concentrations of PAN that were added to the CIMS were quantified simultaneously by a NO y analyzer (TEI 42CY).We also generated PAN using a conventional wet chemistry method by reacting peracetic acid with HNO 3 (Gaffney et al., 1984).The test results from the two PAN sources are consistent, thus this study only shows the result which used the PAN generated from the photolytic source.For tests of PAN in zero air and in relatively clean ambient air at a coastal site (Hok Tsui), a measurable interference from PAN was observed at 62 amu.The result in zero air (i.e., the response of 62 amu signal to the PAN standard) indicates 23 ± 4 pptv of apparent NO 3 + N 2 O 5 signal per ppbv of PAN (one example is shown in Fig. 7).Adding PAN to ambient air at the coastal site yielded similar results.Table 2 shows the results from four tests conducted with varying ambient pollution levels and meteorological conditions.Introducing 5.6-6.(Veres et al., 2008;Roberts et al., 2010).This appeared not to be the case in our CIMS as additions of both PAN and HNO 3 didn't lead to any increase at 62 amu compared to the addition of PAN alone (see Fig. 8).
A surprising finding was much larger interferences at 62 amu when the same spike tests were conducted at the PolyU site, which has very high NO x concentrations, and the interference appeared to increase with ambient NO 2 suggesting that reaction between PAN and NO 2 leads to significant interference at 62 amu.To confirm this, a series of tests were conducted by mixing different levels of NO 2 with synthetic PAN (see an example in Fig. 9). Figure 10 shows the signal at 62 amu as a function of PAN and NO 2 concentrations in zero air, which clearly shows that the interference increases with both PAN and NO 2 .For example, at 5 ppbv of PAN, adding 60 ppbv of NO 2 produces a 400 ppt equivalent N 2 O 5 /NO 3 signal, compared to 150 pptv without NO 2 , indicating an amplifying effect of NO 2 on the previously reported PAN interference.The exact chemical reaction that leads to the interference is not clear.We believe that interference of PAN + NO 2 is related to the thermal dissociation of PAN followed by radical reactions with NO 2 in the heated inlet.The radical reactions might serve as a source of NO 3 and N 2 O 5 , or the products could react with I − to produce NO − 3 ions which is detected at 62 amu.We will evaluate the possible contribution of PAN + NO 2 to the observed high daytime N 2 O 5 + NO 3 for the 2010 autumn study in Sect. 5.

Interference of HNO 3
Another possible source of interference to the NO − 3 ion is HNO 3 since it has a NO 3 group.Several studies have proposed the production of NO − 3 from ion-molecule reactions involving HNO 3 , with somewhat inconsistent results.Huey and co-workers showed that the reaction of I − with HNO 3 (producing NO − 3 ) was rather inefficient.Thornton and coworkers on the other hand found a significant background signal (10-50 Hz) at the NO − 3 mass in a cold iodide CIMS under conditions of long ion-molecule reaction time, and attributed this to the reaction of HNO 3 with I − (Thornton et al., 2003;Thornton and Abbatt, 2005).Roberts and coworkers reported that the HNO 3 was sensitive to the acetate ions and react to produce NO − 3 ions at the 62 amu et al., 2008;Roberts et al., 2010).
The relatively low background signals by adding NO in our study suggest insignificant interference from HNO 3 to our CIMS, which has been corrected during data reduction.This was further confirmed by the addition of HNO 3 .The test was carried out three times for varying ambient conditions to check the repeatability of the results, which are listed in Table 2 with an example being shown in Fig. 8.It can be seen that after introducing a trace amount of HNO 3 to the inlet tube of our TD-CIMS, there was no significant increase in the NO − 3 signal compared to the ambient air.Additionally, adding HNO 3 to a trace level of PAN which resulted in an acetate ion signal of ∼ 0.9 × 10 4 Hz, the NO − 3 signal showed no apparent change compared to that for only adding PAN.These results suggest that the HNO 3 itself and its mixture with acetate ions have no significant interference to the detection of N 2 O 5 + NO 3 via the NO − 3 ion in our TD-CIMS.It should be noted that the result of non-reactivity   of HNO 3 to the acetate ion in our TD-CIMS is different from that obtained by Roberts and coworkers using a NI-PT-CIMS (negative-ion proton-transfer CIMS) (Veres et al., 2008;Roberts et al., 2010), possibly due to different configurations and operation conditions.

Interference of other possible compounds
ClONO 2 , BrONO 2 and HO 2 NO 2 also react efficiently with I − to produce NO − 3 (Huey et al., 1995;Hanson et al., 1996;Zhang et al., 1997;Amelynck et al., 2001).Again, the relatively low instrument background indicated small interference from ClONO 2 and BrONO 2 and any interference from them would have been corrected in our measurements because they cannot be removed by adding NO.The zeroing would not work for HO 2 NO 2 .However, it is impossible for HO 2 NO 2 to pass through the heated inlet tube in our TD-CIMS considering its thermally unstable nature.
In summary, according to the above tests and discussions, reactions between PAN and NO 2 in the heated inlet are found to have significant interference to signal at 62 amu in our TD-CIMS, while HNO 3 and other compounds are not believed to have contributed to signal at 62 amu, although our tests are not exhaustive for including all other chemicals.11).The remaining daytime signal after correction may be a real contribution from N 2 O 5 and NO 3 , but interferences from other untested chemicals (e.g., organic nitrates in addition to PAN) cannot be ruled out.As interferences at the 62 amu channel are large and variable, and some of them may not have been identified in our tests, it is difficult to determine the fraction of real signal from N 2 O 5 + NO 3 in our study at 62 amu.
To compare the observed signals with the concentrations predicted by known chemistry, Fig. 12 shows the source strength and loss frequency of NO 3 , the predicted N 2 O 5 + NO 3 levels for the six polluted episodes, together with corrected N 2 O 5 + NO 3 data (i.e., subtracting the interference due to PAN + NO 2 ).In brief, the reaction of NO 3 with NO is the dominant loss pathway at the study site.In most cases, the NO concentration was relatively high, and the calculated values using steady-state assumption (see SI) were much lower than the observed values (after correction for interference due to PAN + NO 2 ).There is one exception in the late afternoon on 24 October.At 16:00 on this day when O 3 , NO 2 and NO concentrations were 96.concentration (5 min average) could reach 321 pptv if all major removal pathways are considered.In comparison, the measured N 2 O 5 + NO 3 concentration was 286.5 pptv (with correction for PAN + NO 2 interference).The agreement between observation and prediction under the condition of very low NO in this case indicates that the elevated N 2 O 5 in daytime is possible at the study site.The significant discrepancy in most cases may be attributed to other unknown measurement interference (e.g., chemical interference from organic nitrates in addition to PAN), unknown NO 3 source(s) (e.g., chemical reactions between Criegee intermediates and NO 2 suggested by recent studies (Welz et al., 2012;Ouyang et al., 2013), and invalidity of the steady-state assumption for air masses receiving fresh injection of NO).Further studies are required.

Other evidence of daytime NO 3 + N 2 O 5 in Hong Kong
To check the measurement results at 62 amu, we attempted to measure ambient N 2 O 5 with a cold version of CIMS immediately after the present campaign at the same site.The CIMS was configured similar to that described by Kercher et al. (2009).When an unheated inlet tube is used, the reactions of N 2 O 5 with I − produce both the NO − 3 and I(N 2 O 5 ) − cluster ions.The I(N 2 O 5 ) − ion (at 235 amu) is thought to be free from the chemical interferences that can perturb the NO − 3 ion, and thus provides a better measure of N 2 O 5 (Kercher et al., 2009).However, the I(N 2 O 5 ) − sensitivity is influenced by the abundance of the water content in the sample flow.Similar to the setup with the heated inlet and the detection at 62 amu by TD-CIMS, signal at 235 amu with the cold CIMS also showed a daytime peak.Figure 13 gives an example of the ambient I(N 2 O 5 ) − signal (in Hz) and related trace gases taken on 20 December 2010.The 235 amu signal exhibited generally good correlation with the 62 amu signal during both daytime and nighttime (figure not shown here).Calibration of the cold CIMS with the N 2 O 5 standard in dry synthetic air gave a sensitivity of 0.55±0.003Hz pptv −1 .But the true sensitivity for ambient air with moderately high humidity during study is believed to be larger, possibly by a factor of two or more, according to our recent lab tests.
Another independent piece of evidence for the daytime N 2 O 5 + NO 3 in Hong Kong is concurrent increase in the mixing ratios of ClNO 2 (a product of N 2 O 5 hydrolysis) observed in a follow-up study in western Hong Kong (Tung Chung; see SI for the experiment information).At this site, elevated N 2 O 5 + NO 3 signals were also found at 62 amu at daytime during photochemical episodes, with the ClNO 2 signals (208 amu) showing concurrent increases.Figure 14 shows an example for 28 August 2011.On that day, apparent N 2 O 5 + NO 3 inferred from 62 amu exhibited an afternoon peak of 670 pptv (uncorrected 5 min average), and ClNO 2 had a concurrent enhancement to 120 pptv.To estimate the levels of N 2 O 5 that would be needed to sustain such amount of ClNO 2 , we assumed a photostationary steady state for ClNO 2 in the afternoon with an uptake coefficient of 0.03 for N 2 O 5 hydrolysis on aerosol surfaces and a ClNO 2 yield of 10 % with consideration of relatively low content of sea salt in aerosols at the site.The photolysis rate of ClNO 2 was estimated as 7.8 × 10 −4 s −1 using the method by Simon et al. (2009), and the aerosol surface area was 979 mm 2 m −3 from concurrent measurements of aerosol size  distribution.The calculation showed that to produce the observed 120 pptv of ClNO 2 , at least 518 pptv of N 2 O 5 would be required.This result, although is not an absolute proof, supports the possibility of elevated daytime concentrations of N 2 O 5 + NO 3 in Hong Kong.On this day the peak concentration of ClNO 2 was ∼ 600 pptv at night and the ClNO 2 concentration exhibited good correlation (R 2 = 0.78) with the observed NO − 3 signals at 62 amu (figures not shown here).

Conclusions
A TD-CIMS technique, which has been previously applied to field measurements in the US, was deployed to measure ambient NO 3 and N 2 O 5 in urban Hong Kong in a photochemical season.Surprisingly, large signals of NO 3 + N 2 O 5 inferred from 62 amu were frequently observed in daytime, which is in contrast to our current understanding of reactive nitrogen chemistry.Subsequent laboratory and field tests provide new insights into chemical interferences in the TD-CIMS.In particular, we have discovered that reaction between NO 2 and PAN can amplify the interference from PAN at 62 amu.This interference could have contributed 30-50 % to the average daytime NO 3 and N 2 O 5 at our site.It is possible that the additional interferences from untested chemicals may have also contributed to the daytime signal at 62 amu.On the other hand, evidence exists to suggest that the elevated daytime signals at 62 amu may be in part due to real contribution from NO 3 or N 2 O 5 .In viewing of the large and variable interferences at 62 amu and difficulty in correcting them, we conclude that it is not suitable to use the TD-CIMS to measure NO 3 or N 2 O 5 at 62 amu in a high NO x environment like the present study site.Adoption of either 235 amu with a cold  inlet or an optics-based technique is recommended.We also suggest more studies to examine the abundance of daytime NO 3 and N 2 O 5 in similar urban areas with co-existence of high ozone and NO 2 .

Fig. 3 .
Fig. 3. Mass spectra of ambient air in urban Hong Kong at day and night by iodide TD-CIMS with heating the inlet tube to 180 • C.

Fig. 8 .
Fig. 8. Changes of raw signals of CH 3 C(O)O − , NO − 3 , C 3 H 7 C(O)O − , I(H 2 O) − detected by I − TD-CIMS when adding trace amounts of PAN, HNO 3 , or their mix to the ambient air.
3 ppb of PAN to the ambient air resulted in an increase in the NO − 3 signal of 42-75 pptv of NO 3 + N 2 O 5 in our TD-CIMS, again indicating interference from PAN to the field NO 3 /N 2 O 5 measurements via the NO − 3 by TD-CIMS.The mechanism for the interference of PAN on the NO − 3 signal is unclear.Some researchers proposed that the CH 3 C(O)O − ion (produced from the reaction of I − with CH 3 C(O)O 2the product of PAN thermo-dissociation) likely reacts with HNO 3 to produce NO − 3

Fig. 9 .Fig. 9 .
Fig. 9. Fig. 9. Changes of signals at 59 and 62 amu when NO 2 was added to the sample flow of synthetic PAN. 34

Fig. 10 .
Fig. 10.Interference of NO 2 to equivalent NO − 3 concentration under different levels of PAN.Vertical bars are standard deviations of tests results.

5
Contribution of interference to daytime concentrationsAs indicated above, laboratory and field tests revealed significant interference from PAN and NO 2 to 62 amu in our TD-CIMS.The contribution from this interference to the observed daytime N 2 O 5 + NO 3 in this study was examined and corrected based on the tests results in Fig.10.During the field campaign at PolyU, PAN was concurrently measured by the same TD-CIMS and a GC-ECD instrument (gas chromatography with an electron capture detector).For the six episodes shown in Fig.6, the interference would contribute to 41.3-67.0%, 29.0-39.8%, 32.2-73.2%, 25.6-49.2%, 19.4- 43.4 %, 20.4-77.8  % of the daytime signals in the 6 episodes (two examples on 23 October and 30 November are shown in Fig.

Fig. 12 .
Fig. 12. NO 3 production rate from NO 2 and ozone, loss frequency of NO 3 from NO reaction, photolysis, VOC oxidation and N 2 O 5 hydrolysis, steady-state calculated N 2 O 5 + NO 3 and observed N 2 O 5 + NO 3 after correction (for the interference from PAN + NO 2 ) for the six pollution episode days.

Fig. 13 .
Fig. 13.Time series of I(N 2 O 5 ) − signal at 235 amu, solar radiation and relative humidity in urban Hong Kong on 20 December 2010.

Fig. 14 .
Fig. 14.A case showing concurrent increase of ClNO 2 and N 2 O 5 at daytime during a photochemical episode at a suburban site (Tung Chung) in western Hong Kong.

Table 1 .
Ion-molecule reactions with product of NO − 3 and the rate constants.

Table 2 .
Summary of the interference tests for the TD-CIMS.The concentrations of PAN and HNO 3 that were added to the ambient air were determined by a NO y analyzer. *