On the Determination of Continuous Submicrometer Liquid Aerosol-Size Distributions with Low Pressure Impactors
Titel:
On the Determination of Continuous Submicrometer Liquid Aerosol-Size Distributions with Low Pressure Impactors
Auteur:
Kauppinen, Esko I.
Verschenen in:
Aerosol science and technology
Paginering:
Jaargang 16 (1992) nr. 3 pagina's 171-197
Jaar:
1992
Inhoud:
The aspects associated with the determination of continuous submicrometer aerosol-size distributions using multijet low pressure impactors have been studied. Multiple sets of error-free and noisy, simulated data sets have been inverted, and impactors have been compared with the differential mobility particle-size analysis (DMA) method by using well-defined, laboratory-generated liquid oleic acid aerosols tagged with ammonium fluorescein. Impactors included in this study were a Berner-type impactor HAUKE 25/0.015 (BLPI), a modified University of Washington Mark 5 impactor (KLPI), and the impactor designed at the University of Florida (LLPI). The inversion of simulated error-free impactor data (i.e., the data with perfect kernel functions) for unimodal submicrometer aerosols with a small (2.5%) stage mass error estimate yields results very close to input distributions, when the method based on constrained regularization is used in the inversion. When the error estimate is increased, inverted spectra are flattened. However, they remain clearly unimodal. When normally distributed random error is added to the data and the error estimate for each data point equals the standard deviation of the random error, the fraction of bimodal and trimodal inverted spectra increases with a rise in the random error level and with the asymmetricity of the kernel functions. When the random error level and data error estimates are equal to or smaller than 10%, inverted spectra are mainly unimodal close to input distribution for both error-free and noisy data. The inversion of impactor data from the detailed laboratory experiments (i.e., the data with real kernel functions) indicates that only BLPI kernel functions are accurate enough to yield unimodal distributions close to those measured with the DMA. When the stage mass error estimate is increased beyond the stage mass determination error, unimodal spectra also for the KLPI and LLPI are found. The decrease of the BLPI stage mass error estimate below the experimental error increases the agreement with DMA results. In most cases the error estimate for BLPI stage masses can be decreased to 2.5%, indicating the validity of both BLPI submicrometer kernel functions and the fluorometric method used to determine stage mass concentrations.The aspects associated with the determination of continuous submicrometer aerosol-size distributions using multijet low pressure impactors have been studied. Multiple sets of error-free and noisy, simulated data sets have been inverted, and impactors have been compared with the differential mobility particle-size analysis (DMA) method by using well-defined, laboratory-generated liquid oleic acid aerosols tagged with ammonium fluorescein. Impactors included in this study were a Berner-type impactor HAUKE 25/0.015 (BLPI), a modified University of Washington Mark 5 impactor (KLPI), and the impactor designed at the University of Florida (LLPI). The inversion of simulated error-free impactor data (i.e., the data with perfect kernel functions) for unimodal submicrometer aerosols with a small (2.5%) stage mass error estimate yields results very close to input distributions, when the method based on constrained regularization is used in the inversion. When the error estimate is increased, inverted spectra are flattened. However, they remain clearly unimodal. When normally distributed random error is added to the data and the error estimate for each data point equals the standard deviation of the random error, the fraction of bimodal and trimodal inverted spectra increases with a rise in the random error level and with the asymmetricity of the kernel functions. When the random error level and data error estimates are equal to or smaller than 10%, inverted spectra are mainly unimodal close to input distribution for both error-free and noisy data. The inversion of impactor data from the detailed laboratory experiments (i.e., the data with real kernel functions) indicates that only BLPI kernel functions are accurate enough to yield unimodal distributions close to those measured with the DMA. When the stage mass error estimate is increased beyond the stage mass determination error, unimodal spectra also for the KLPI and LLPI are found. The decrease of the BLPI stage mass error estimate below the experimental error increases the agreement with DMA results. In most cases the error estimate for BLPI stage masses can be decreased to 2.5%, indicating the validity of both BLPI submicrometer kernel functions and the fluorometric method used to determine stage mass concentrations.The aspects associated with the determination of continuous submicrometer aerosol-size distributions using multijet low pressure impactors have been studied. Multiple sets of error-free and noisy, simulated data sets have been inverted, and impactors have been compared with the differential mobility particle-size analysis (DMA) method by using well-defined, laboratory-generated liquid oleic acid aerosols tagged with ammonium fluorescein. Impactors included in this study were a Berner-type impactor HAUKE 25/0.015 (BLPI), a modified University of Washington Mark 5 impactor (KLPI), and the impactor designed at the University of Florida (LLPI). The inversion of simulated error-free impactor data (i.e., the data with perfect kernel functions) for unimodal submicrometer aerosols with a small (2.5%) stage mass error estimate yields results very close to input distributions, when the method based on constrained regularization is used in the inversion. When the error estimate is increased, inverted spectra are flattened. However, they remain clearly unimodal. When normally distributed random error is added to the data and the error estimate for each data point equals the standard deviation of the random error, the fraction of bimodal and trimodal inverted spectra increases with a rise in the random error level and with the asymmetricity of the kernel functions. When the random error level and data error estimates are equal to or smaller than 10%, inverted spectra are mainly unimodal close to input distribution for both error-free and noisy data. The inversion of impactor data from the detailed laboratory experiments (i.e., the data with real kernel functions) indicates that only BLPI kernel functions are accurate enough to yield unimodal distributions close to those measured with the DMA. When the stage mass error estimate is increased beyond the stage mass determination error, unimodal spectra also for the KLPI and LLPI are found. The decrease of the BLPI stage mass error estimate below the experimental error increases the agreement with DMA results. In most cases the error estimate for BLPI stage masses can be decreased to 2.5%, indicating the validity of both BLPI submicrometer kernel functions and the fluorometric method used to determine stage mass concentrations.The aspects associated with the determination of continuous submicrometer aerosol-size distributions using multijet low pressure impactors have been studied. Multiple sets of error-free and noisy, simulated data sets have been inverted, and impactors have been compared with the differential mobility particle-size analysis (DMA) method by using well-defined, laboratory-generated liquid oleic acid aerosols tagged with ammonium fluorescein. Impactors included in this study were a Berner-type impactor HAUKE 25/0.015 (BLPI), a modified University of Washington Mark 5 impactor (KLPI), and the impactor designed at the University of Florida (LLPI). The inversion of simulated error-free impactor data (i.e., the data with perfect kernel functions) for unimodal submicrometer aerosols with a small (2.5%) stage mass error estimate yields results very close to input distributions, when the method based on constrained regularization is used in the inversion. When the error estimate is increased, inverted spectra are flattened. However, they remain clearly unimodal. When normally distributed random error is added to the data and the error estimate for each data point equals the standard deviation of the random error, the fraction of bimodal and trimodal inverted spectra increases with a rise in the random error level and with the asymmetricity of the kernel functions. When the random error level and data error estimates are equal to or smaller than 10%, inverted spectra are mainly unimodal close to input distribution for both error-free and noisy data. The inversion of impactor data from the detailed laboratory experiments (i.e., the data with real kernel functions) indicates that only BLPI kernel functions are accurate enough to yield unimodal distributions close to those measured with the DMA. When the stage mass error estimate is increased beyond the stage mass determination error, unimodal spectra also for the KLPI and LLPI are found. The decrease of the BLPI stage mass error estimate below the experimental error increases the agreement with DMA results. In most cases the error estimate for BLPI stage masses can be decreased to 2.5%, indicating the validity of both BLPI submicrometer kernel functions and the fluorometric method used to determine stage mass concentrations.
Tijdschrift beschrijving