1.  Does the paper address relevant scientific questions within the scope of AMT? Yes
2. Does the paper present novel concepts, ideas, tools, or data? Yes
3. Are substantial conclusions reached? Partly, see major points of criticism.
4. Are the scientific methods and assumptions valid and clearly outlined? Partly, see major points of criticism.
5. Are the results sufficient to support the interpretations and conclusions? Partly, see major points of criticism.
6. Is the description of experiments and calculations sufficiently complete and precise to allow their reproduction by fellow scientists (traceability of results)? Partly, see major points of criticism.
7. Do the authors give proper credit to related work and clearly indicate their own new/original contribution? Yes
8. Does the title clearly reflect the contents of the paper? Yes
9. Does the abstract provide a concise and complete summary? Yes
10. Is the overall presentation well-structured and clear? Yes
11. Is the language fluent and precise? Yes
12. Are mathematical formulae, symbols, abbreviations, and units correctly defined and used? Overall, yes
13. Should any parts of the paper (text, formulae, figures, tables) be clarified, reduced, combined, or eliminated? Some parts need clarification and additional detail information. See below.
14. Are the number and quality of references appropriate? Yes
15. Is the amount and quality of supplementary material appropriate? Yes

Major points of criticism:

1. Line 55 to 59 describes the technical challenge to which the paper is dedicated: Increasing the detection efficiency of TXRF by complete excitation of the deposit and complete detection of the fluorescence response. The authors present a solution by shrinking the lateral diameters of centric deposit patterns on cascade substrates (carriers) to less than 5 mm and describe an appropriate cascade impactor design. To the reader the 5 mm criterion seems to be an arbitrary choice and a justification is lacking. A justification could e. g. be derived from an excitation beam profile analysis along with a description of the detector’s aperture and lateral efficiency distribution. Both would, e. g. for the used Bruker TStar spectrometer, clearly reveal that other than the authors implicitly assume even within the “5 mm area” the excitation and detection would have radial dependencies, i.e., decrease with distance to the center.
2. In line 120 authors describe a Reynolds number of ~3000 in the nozzles as a design criterion. This is in my understanding for cylindrical pipes the transition regime between turbulent (>3000) and laminar (<=2300) flow. In line 125 to 130 authors describe that the flow profile in the nozzles should be “plug-shaped”, i. e. turbulent, rather than “parabolic” aka laminar. This is in contradiction to line 137, where a Reynolds number of < 3000 is described as desirable. The relation between number and diameter of nozzles and mean velocity could be described more precisely and as a design challenge. It is desirable to have the finally applied ratios of nozzle lengths to diameters presented as a result of theoretical calculations or CFD modelling.
3. Attrition (section 3.3.1). To me, sample contamination by metallic attrition appears a bit far-fetched. This section could possibly be eliminated. Authors could e. g. have made a point in discussing attrition by solid, abrasive particles. The authors also did not consider the possible contamination of a quartz disc by unintended mechanical contact with metallic parts (e. g. impactor structure, tweezers…) or other materials which have had mechanical contact to metals. This deems to me a more likely possible source of contamination. The presentation of related results in section 4, Table 2 has some deficiencies as will be discussed below.
4. Line 265: The statement on the “effective area” is vague. As pointed out above, the efficiency will most probably have a radial dependency. Authors do not present an argument why this could be ignored. Line 269: “Calibration samples as external standards”. This is a crucial point in quantitative TXRF analysis, and the paper is much too vague here. The authors should explain the calibration samples in detail. I assume that these were prepared by drying a small drop of an aqueous solution with known element quantity (probably Yttrium?) in the center of the carrier. I have the impression that the authors implicitly assume for the whole deposition pattern area the same detection efficiency as for a punctiform centrical deposition of an internal standard and they should give a proper justification for this assumption. As the commercial Bruker software for the TStar offers several options for quantification the chosen method (internal fundamental parameters, internal standard…) should be mentioned.
5. Chapter 4, Table 2: “…mean blank values of impactor stages 1 to 5 …”: I guess that this are the means over repeated measurements. Lines 291 to 295. I would expect a proper measure of variance, at least an estimation of the uncertainty. Each single measurement should come with an uncertainty provided by the analyzing software. The absolute lower detection limits (LLOD) should be quantified. Line 299 to 301: The significance of the statement should be reevaluated considering realistic uncertainties in the measurements. I am reluctant to accept conclusions without proper consideration of uncertainties, the more so as the data are close to the LLOD.
6. Chapter 4 from line 345 on and Figure 5 and Supplement Table S.2: “Figure 5 shows the mass concentrations of the trace metals Zn, Cu, Mn, Pb and Ni collected during these three sampling periods in three size fractions, i. e. PM1 (top of blue bar), PM2.5 (top of green bar) and PM10 (top of red bar).”. Authors should describe how they calculated the PMx mass fractions from the masses determined by TXRF on the five cascade impactor stages. Were for the 0.11 to 1 µm fraction the masses from stages 3 to 5 just summed up, and accordingly for the other PM fractions in Figure 5? If the coarse (10 µm) stage was included into the calculations at least all values labeled as PM10 (top of red bar) do contain a priori unknown quantities of particles larger than 10 µm as no preseparator was reportedly used and the separation curve steepness according to DIN 481 or US EPA was not determined. However, the photo (Fig. 3) probably shows an inlet separator which is not described at all.
7. Conclusions and Outlook seems not well-balanced: One half of the text can be considered merely as outlook. The following three points can in essence be regarded original results from this work:
   1. “Due to the compact arrangement of the impactor 375 nozzles, the high detection sensitivity of the TXRF analysis method can be fully exploited.” A convincing proof of this statement is however lacking. The statement in lines 375 to 379 is valid.
   2. The statement on the “new spin-coating method” is valid.
   3. No contamination issues during mounting/dismounting of carriers.

Minor points / editorial suggestions:

Line 76 to 77: The citation refers to rather broad outdoor aerosol size distributions,  e. g. accumulation or coarse mode. There might be outdoor scenarios where the Aitken mode is interesting to observe even with high time resolution, or indoor scenarios were the < 0.1 µm fraction is dominant. Authors should comment on that.

Line 82: “slm” (standard litre per minute) refers to us-american nomenclature with standard conditions 0 deg Celsius  (32 deg F) and 1.0125 bar (14.69 psia). European nomenclature uses subscript “_s” for standard flow at 20 C and 1.0125 bar, e.g., m³_s /h. Standard flow at 0 C and 1.0125 bar is in Europe expressed as e.g., m³_n / h with  subscript “_n”

Line 165: “in the low single-digit percentage range” ?? “few percent relative deviation” sounds better.

Statement summary of the reviewer:

Pro: The three design targets described in line 62 were reached. The overall functionality of a newly designed impactor could be demonstrated in the field; the high time resolution as well. The stages design with respect to flow, nozzle numbers, diameters, lengths, and lateral arrangement was comprehensively and in detail described and supported by theoretical considerations. Application of an adhesive on carriers by spin coating was successfully demonstrated and seems to be an interesting alternative to spraying.

Contra: Unfortunately, the paper does not present convincing proof of the expected advantage of the new design over commercial impactors such as increased efficiency of TXRF quantification of deposited elements. It would be desirable if the authors discuss shape and intensity profile of the detection area of the TStar in more detail to underpin their point of view. The discussion on cross contaminations and contamination by attrition lacks proper consideration of measurement uncertainties and LLODs. The quantitative evaluation of outdoor data in terms of element concentrations in PM fractions is not described sufficiently.

This work by Crazzolara and Held developed a new cascade impactor that improves the detection limits of heavy metals in aerosol particles. What’s more, the newly equipment was in small size with low detection limits for metal elements, which is beneficial to monitoring work in field observation. This article is well organized, informative and in line with the scope of AMT. I suggested that the manuscript can be accepted and published after addressing the following concerns.

Main concerns:

1. The author stated that the detection limit of TXRF is superior to XRF (Yoneda and Horiuchi, 1971), and can reach down to a few picograms of absolute mass on the sample carrier substrate (Streli 2006). Can the author give the detailed detection limits of TXRF of some elements for comparison? And more latest references should be provided here.

1. Line 84-85: “…… without exceeding a critical Reynolds number of 3000” and Table 1. It puzzles me that why the criterion set as 3000 in a circular area with a diameter of less than 5 mm on stages 3, 4 and 5? And why small nominal nozzle diameters corresponded to the high Reynolds number?

1. Line 243-245: The operating temperature of the sensor is ranged from -20 to +80 °C. The question is whether the sensor have an applicable relative humidity RH range? For example, in the coastal regions, high humidity and salinity environment may cause damage to the instrument, such as corrosion.

1. Line 251-254: How to set the duration of the sample time to ensure that the sample meets the needs of the analysis? For example, if the collection time is too short in a very clean environment, the sample volume may be not sufficient for the test, but if the collection time is too long, the sample may be overloaded.

1. Line 295: The detection limits of Fe, Cr and Ni by TXRF analysis should be given here.

1. In the section “4.3 Collection of particles in outdoor air”, does the lower mass concentrations of Pb in PM₁₀ (ranging from 1.1 to 1.7 ng m^-3) and Ni in PM₁₀ (ranging from 0.4 to 0.6 ng m^-3) imply that a sampling period of 30 min is not sufficient for analysis.

1. Have the authors compared the capture efficiency of the new cascade impactor with other commercially cascade impactors such as MOUDI Impactor Series (TSI Incorporated, USA) or Andersen Cascade Impactor (Tisch Environmental, Inc. USA)?

Minor concerns:

1. PM₁₀, PM₅ and PM₁ should be subscripted. Many similar issues in the manuscript.

In this study, a new impactor optimised for TXRF analysis was developed, demonstrating that a large number of different heavy metals can be detected and quantified in the PM10, PM2.5 and PM1 size fractions after collection periods of 30 minutes. Overall, the new impactor bears potential to improve the quantification of particulate trace metals and other elements in PM10, PM2.5 and PM1 with high time resolution. Major comments are as follows:

1. In the introduction, it is said that “Despite promising results, commercial impactors are not fully optimised for TXRF analysis: The area on the sample carrier in which the classifying nozzles deposit the particles is usually significantly larger than the area analysed by TXRF.” The inappropriate area was the only limitation? More advances should be showed.
2. Why did the new impactor select PM10, PM2.5 and PM1, but not PM1, PM1-2.5 and PM2.5-10? PM0.11-1, PM1-2.5 and PM2.5-10 were showed in Figure 5, so what the impactor actually sampled?
3. What elements can be detected by the TXRF? Why the Fe, Cr, and Ni mean blank values were showed in section 4.1, but the concentrations of Zn, Cu, Mn, Pb and Ni were showed in section 4.3? How about other metals, such as Co, V, As? Blank values and detection limits of all the measured elements should be summarized.
4. The concentrations can largely changed in the environment. How to set the sampling time to ensure that the sample meets the needs of the analysis and is not overloaded?
5. Measurement results using this impactor and those using commercial impactors should be compared, to show the improvement of this new impactor.