SP-AMS PMF with fullerenes

SP-AMS PMF with fullerenes

James Allan

Zainab Bibi

University of Manchester

The person who did the actual work

• Zainab Bibi (pictured)

• Paper currently in review at ACP:

https://acp.copernicus.org/prepr

ints/acp-2020-890/

• Due to submit thesis in the

coming year

BC apportionment/attribution

• Methods for BC source apportionment include:

• Optical (e.g. Aethalometer, PAS)

• Use of tracers (e.g. levoglucosan, potassium, fossil fuel tracers)

• Isotopic analysis, principally

14

_C

• Existing established methods typically split BC into two sources, e.g.:

• Black vs brown carbon

• Traffic vs wood burning

• Modern vs fossil carbon

• The SP-AMS delivers mass spectral data. Can we do better by applying

PMF?

Fullerenes

• The C

_x

peaks can be grouped into ‘low’ (C

_1-5

), ‘mid’ (C

_6-31

) and ‘high’

(C

_>31

) m/z ranges (Onasch et al., 2012)

• The ‘low’ C

_x

peaks do not show enough variation and the ‘mid’ C

_x

peaks get lost amongst organics in ambient spectra. Can we get better

PMF factorisation through inclusion of ‘high’ C

_x

peaks in the mass

spectral matrix?

• Previous studies have reported fullerenes in the high C

_x

range from

certain soot sources in the SP-AMS (Fortner et al., 2012)

• Not sure whether they are produced in the flame or the instrument

The case study

• Bonfire night, an annual festival on or

around the 5

th

_{of November celebrating the}

gunpowder plot in 1605

• In 2014 we ran a suite of instruments at the

lab in Manchester for that night and a

week either side of it

• Fortuitously, that year a strong nocturnal

inversion formed, presenting a very good

case study of biomass burning and

fireworks against typical wintertime UK

urban background

• The subject of four papers involving the

AMS/Aethalometer (Reyes Villegas et al.,

2018), TOF-CIMS (Priestley et al., 2018),

SP2 (Liu et al., 2017) and now SP-AMS (Bibi

et al., in review)

3000 2000 1000 0 rB C ( c ou nts s -1 ) 29/10/2014 31/10/2014 02/11/2014 04/11/2014 06/11/2014 08/11/2014 10/11/2014 Date and Time (UTC)

12 8 4 0 B C ( µgm -3 ) 600 400 200 0 B rC ( Mm -1 ) 20 15 10 5 0 B C ( µgm -3 ) rBC (HR-SP-AMS) BC(AE31) BrC (AE31) BC (MAAP)

SP-AMS configuration

• The SP-AMS alternated between the standard ‘V’ and ‘W’ modes,

with an extra ‘fullerene’ modes, which was ‘V’ mode with a much

lower pulsing frequency (allowing the study of m/z’s of 1000s)

• ‘W’ mode proved unusable

• The lower m/z channels were analysed with PIKA, whereas UMR was used to

provide the fullerene data

• Was not calibrated for rBC (didn’t have a protocol at the time) – all

data is presented in counts per s

• Also alternated with a catalytic stripper, not presented here (that can

wait for another time…)

Error treatment

• The use of a combination of large and small signals ensured factors were dominated by

splits in the large signals, overcome by application of the ‘model error’ parameter

• Through trial and error, a value of 0.1 was settled on as giving the ‘best’ PMF outputs

• A more rigorous approach such as Corbin et al. (2015) was not possible due to the

combination of PIKA and UMR data

1200 1000 800 600 400 200 0 S NR 800 600 400 200 m/z C CH CH2 CH3 O CH4 HO H2O CO2plus2 C2 C2H C2H2 CHN C2H3 CO CHO CH3NC2H5 CH2O CH4NC2H6 CH3O C3 C3H C3H2 C3H3 C3H4 C2HO CHN2 C2H3NC3H5 C2H2O C2H4N C3H6 CHNO C2H3O C3H7 CO2 C2H4O C2H6N C3H8 CHO2 CH3NO_C2H5O C2H7N CH2S CH2O2 CH4NOC2H6OCH3SC4 C4H C4H2 C4H3 C4H4 C3HO C3H3N C4H5 C3H2O C4H6 C3H3O C4H7 C2O2 C3H4O C4H8 C2HO2 C3H5O C3H7N C4H9 C2H2O2

C2H4NO_C2H3O2C3H6OC4H10_C2H3SCOP CH3N2O C2H5NOC3H7O CSO C5 CH2NO2 C2H4O2 C3H8O CHSO C5H C2H5O2 CH2SO C5H2 C2H6O2 CH3SOC5H3 C4O C5H4 C4HO C5H5 C4H2O C5H6 C4H3O C5H7 C3O2 C4H4O C5H8 C3HO2 C4H5O C5H9 C3H2O2 C4H6O C5H10 C3H3O2 C3H5NO C4H7O C5H11 C6 C2H2NO2C3H4O2 CH4N4 C4H8OC5H12_C6H C3H5O2 C2H5N2O C3H7NO C4H9O C6H2 C3H6O2 C4H10O C6H3 C3H7O2 CS2 C5O C6H4 C3H8O2 C5HOC6H5 C3H9O2 CH2SO2 C5H2OC6H6 C5H3O C6H7 C4O2 C5H4OC6H8 C4HO2 C5H5O C6H9 C4H2O2 C5H6O C6H10 C4H3O2 C5H7O C6H11 C7 C4H4O2C5H8O C6H12C7H C4H5O2 C5H9O C6H13 C7H2

C3H4NO2_C4H6O2C5H10OC3H3O3C6H14C6HNC7H3 C3H5NO2 C4H7O2 C4H9NO C5H11OC6O C2H2NO3C7H4 C3H6NO2 C4H8O2 C5H12O C2HSO2C6HOC7H5 C3H7NO2 C4H9O2 C2H2SO2_C6H2O C2H4NO3C7H6 C4H10O2 C6H3OC7H7 C5O2 C2H4O4C6H4OC7H8 C5HO2 C6H5O C7H9 C5H2O2 C6H6O C6H8N C7H10CH4Br C5H3O2C6H7O C7H11 C8 C5H4O2C6H8O C7H12C8H C5H5O2 C6H9O C6H11N C7H13 C8H2 C5H6O2 C6H10O C7H14 C3H3N2O2_C8H3 C5H7O2 C6H11O C6H13N C7H15 C7O C8H4 C4H6NO2 C5H8O2

C4H8N2OC6H12OC7H16C7HOC8H5

C4H7NO2 C5H9O2 C6H13O C7H2OC8H6 C5H10O2 C6H14OC7H3OC8H7 C5H11O2 C6O2 C7H4O C8H8 C6HO2 C7H5O C8H9 C6H2O2 C7H6O C8H10 C6H3O2 C7H7O C8H11_C9 C6H4O2C7H8O C8H12_C9H C6H5O2 C7H9OC8H13 C9H2 C6H6O2 C7H10O C8H14C9H3 C6H7O2 C7H11OC8H15 C8O C9H4 C6H8O2 C7H12OC6H9O2C7H13O_C5H6O3C8H16C8H2OC8H17C8OH_C9H5 C4H6N2O2C6H10O2_C6H14N2C7H14OC9H6

C8H18 C8H3O C4H5NO3C9H7 C6H11O2 C7H15O C4H6NO3_C5H8O3 C56 C57C58C59C60C61C62C63C64C65C66C67C68C69C70C71C72C73C74C75C76C77C78C79C80C81C82C83 0.2 < SNR < 2 "weak" variable SNR < 0.2 "bad" variable 15 10 5 0 S N R 800 600 400 200 m/z C CH CH2 CH3 O CH4 HO H2O CO2plus2 C2 C2H C2H2 CHN C2H3 CO CHO CH3N C2H5 CH2O CH4N C2H6 CH3O C3 C3H C3H2 C3H3 C3H4 C2HO CHN2 C2H3N C3H5 C2H2O C2H4N C3H6 CHNO C2H3O C3H7CO2 C2H4OC2H6N C3H8 CHO2 CH3NO C2H5O C2H7N CH2S CH2O2 CH4NO C2H6O CH3S C4 C4H C4H2 C4H3C4H4 C3HO C3H3N C4H5 C3H2O C4H6 C3H3OC4H7C2O2 C3H4O C4H8 C2HO2 C3H5O C3H7N C4H9 C2H2O2 C2H4NO C3H6O C4H10 COP C2H3S C2H3O2 CH3N2O C2H5NO C3H7O CSO C5 CH2NO2 C2H4O2 C3H8O CHSO C5H C2H5O2 CH2SO C5H2 C2H6O2 CH3SO C5H3 C4O C5H4 C4HO C5H5 C4H2O C5H6 C4H3O C5H7 C3O2 C4H4O C5H8 C3HO2 C4H5O C5H9 C3H2O2 C4H6O C5H10 C3H3O2 C3H5NO C4H7O C5H11C6 C2H2NO2 C3H4O2 CH4N4 C4H8O C5H12 C6H C3H5O2 C2H5N2O C3H7NO C4H9O C6H2 C3H6O2C4H10O C6H3 C3H7O2 CS2 C5O C6H4 C3H8O2 C5HO C6H5 C3H9O2 CH2SO2 C5H2O C6H6 C5H3O C6H7 C4O2 C5H4O C6H8 C4HO2 C5H5O C6H9 C4H2O2 C5H6O C6H10 C4H3O2 C5H7O C6H11C7 C4H4O2 C5H8OC6H12C7H C4H5O2 C5H9O C6H13C7H2 C3H4NO2 C4H6O2 C5H10O C6H14 C3H3O3 C6HN C7H3 C3H5NO2 C4H7O2 C4H9NO C5H11O C6O C2H2NO3 C7H4 C3H6NO2 C4H8O2 C5H12O C2HSO2 C6HO C7H5 C3H7NO2 C4H9O2 C2H2SO2 C6H2O C2H4NO3 C7H6 C4H10O2 C6H3O C7H7 C5O2 C2H4O4 C6H4O C7H8 C5HO2 C6H5O C7H9 C5H2O2 C6H6O C6H8N C7H10 CH4Br C5H3O2 C6H7O C7H11C8 C5H4O2 C6H8O C7H12C8H C5H5O2 C6H9O C6H11N C7H13 C8H2 C5H6O2 C6H10O C7H14 C3H3N2O2 C8H3 C5H7O2 C6H11O C6H13N C7H15 C7O C8H4 C4H6NO2 C5H8O2 C4H8N2O C6H12O C7H16 C7HO C8H5 C4H7NO2 C5H9O2 C6H13O C7H2O C8H6 C5H10O2 C6H14OC7H3O C8H7 C5H11O2 C6O2 C7H4O C8H8 C6HO2 C7H5O C8H9 C6H2O2 C7H6O C8H10 C6H3O2 C7H7O C8H11_C9 C6H4O2 C7H8O C8H12 C9H C6H5O2 C7H9OC8H13C9H2 C6H6O2 C7H10O C8H14C9H3 C6H7O2 C7H11O_C8H15 C8O C9H4 C6H8O2 C7H12O C8H16 C8OH C9H5 C6H9O2 C7H13O C8H17 C8H2O C5H6O3 C4H6N2O2C9H6 C6H10O2_C7H14O C6H14N2 C8H18 C8H3O C4H5NO3 C9H7 C6H11O2 C7H15O C4H6NO3 C5H8O3 C56 C57 C58 C59 C60 C61 C62 C63 C64 C65 C66 C67 C68 C69 C70 C71 C72 C73 C74 C75 C76 C77 C78 C79 C80 C81 C82 C83 0.2 < SNR < 2 "weak" variable SNR < 0.2 "bad" variable 10 8 6 4 2 0 S N R 800 600 400 200 m/z C CH CH2 CH3_O CH4 HO H2O CO2plus2 C2 C2H C2H2 CHN C2H3_CHOCO CH3N C2H5 CH2O CH4N C2H6 CH3O C3 C3H C3H2 C3H3C3H4 C2HO CHN2 C2H3N C3H5 C2H2O_C2H4N C3H6 CHNO C2H3O C3H7CO2 C2H4O C2H6N C3H8 CHO2 CH3NO C2H5O C2H7N CH2S CH2O2 CH4NO C2H6O CH3S C4 C4H C4H2 C4H3 C4H4 C3HO C3H3N C4H5 C3H2O C4H6 C3H3OC4H7C2O2 C3H4O C4H8 C2HO2 C3H5O C3H7N_C4H9 C2H2O2 C2H4NO C3H6O C4H10 COP C2H3S C2H3O2 CH3N2O C2H5NO C3H7O CSO C5 CH2NO2 C2H4O2 C3H8O CHSO C5H C2H5O2 CH2SO C5H2 C2H6O2 CH3SO C5H3 C4O C5H4 C4HO C5H5 C4H2O C5H6 C4H3O_C5H7C3O2 C4H4O C5H8 C3HO2 C4H5OC5H9 C3H2O2 C4H6O C5H10 C3H3O2 C3H5NO C4H7O C5H11C6 C2H2NO2 C3H4O2 CH4N4 C4H8O C5H12 C6H C3H5O2 C2H5N2O C3H7NO C4H9O C6H2 C3H6O2 C4H10O C6H3 C3H7O2 CS2 C5O C6H4 C3H8O2 C5HO C6H5 C3H9O2 CH2SO2 C5H2O C6H6 C5H3O C6H7 C4O2 C5H4O C6H8 C4HO2 C5H5O C6H9 C4H2O2 C5H6O C6H10 C4H3O2 C5H7O C6H11C7 C4H4O2 C5H8OC6H12C7H C4H5O2 C5H9O_C6H13C7H2 C3H4NO2 C4H6O2 C5H10O C6H14 C3H3O3 C6HN C7H3 C3H5NO2 C4H7O2 C4H9NO C5H11O C6O C2H2NO3 C7H4 C3H6NO2 C4H8O2 C5H12O C2HSO2 C6HO C7H5 C3H7NO2 C4H9O2 C2H2SO2 C6H2O C2H4NO3 C7H6 C4H10O2 C6H3O C7H7 C5O2 C2H4O4 C6H4O C7H8 C5HO2 C6H5O C7H9 C5H2O2 C6H6O C6H8N C7H10 CH4Br C5H3O2 C6H7OC7H11C8 C5H4O2 C6H8O C7H12C8H C5H5O2 C6H9O C6H11N C7H13 C8H2 C5H6O2 C6H10O_C7H14 C3H3N2O2 C8H3 C5H7O2 C6H11O C6H13N C7H15 C7O C8H4 C4H6NO2 C5H8O2 C4H8N2O C6H12O C7H16 C7HO C8H5 C4H7NO2 C5H9O2 C6H13O C7H2O C8H6 C5H10O2 C6H14OC7H3O C8H7 C5H11O2 C6O2 C7H4O C8H8 C6HO2 C7H5O C8H9 C6H2O2 C7H6O C8H10 C6H3O2 C7H7O C8H11C9 C6H4O2 C7H8OC8H12C9H C6H5O2 C7H9OC8H13C9H2 C6H6O2 C7H10O C8H14C9H3 C6H7O2 C7H11O_C8H15C8O C9H4 C6H8O2 C7H12O C8H16C8OH C9H5 C6H9O2 C7H13O_C8H17 C8H2O C5H6O3 C4H6N2O2C9H6 C6H10O2_C7H14O C6H14N2 C8H18 C8H3O C4H5NO3 C9H7 C6H11O2 C7H15O C4H6NO3 C5H8O3 C56C57C58C59C60C61C62C63C64C65C66C67C68C69C70C71C72C73C74C75C76C77C78C79C80C81C82C83 0.2 < SNR < 2 "weak" variable SNR < 0.2 "bad" variable

5 factor solution

• Fullerenes present in 2 factors, both associated with bonfire

emissions

• Fullerenes not present in traffic or domestic BBOA

• Only one ambiguous factor, one with HULIS (associated with both

bonfire and other sources

8000

6000

4000

2000

0

05/11/2014

07/11/2014

09/11/2014

Date and Time (UTC)

600

400

200

0

Coun

ts

/s

3000

2000

1000

0

Coun

ts

/s

60

40

20

0

x

10

3

Non-Bonfire Night Factors (3, 4)

Factor 3 = Domestic Burning

Factor 4 = Hydrocarbon-like OA

Bonfire Night Factors (1, 2 and 5)

Factor1 = BC and HULIS

Factor5 = Fullerenes

Factor2 = BBOA

(a)

(b)

0.20

0.15

0.10

0.05

0.00

F

rac

ti

on

of

s

ign

al

100

80

60

40

20

0

30

20

10

0

x

10

-3

0.12

0.10

0.08

0.06

0.04

0.02

0.00

80

60

40

20

0

x

10

-3

0.12

0.08

0.04

0.00

8

6

4

2

0

x

10

-6

1000

900

800

700

3.0

2.0

1.0

0.0

x

10

-6

2.5

2.0

1.5

1.0

0.5

0.0

x

10

-6

3.0

2.0

1.0

0.0

x

10

-6

60

40

20

0

x

10

-6

Factor 5 = Fullerenes

Factor 4 = Hydrocarbon-Like OA

Factor 3 = Domestic burning

Factor 2 = Biomass Burning OA

Factor 1 = BC and HULIS

44

H:C = 1.13

O:C = 0.76

H:C = 1.74

O:C = 0.03

H:C = 1.67

O:C = 0.31

H:C = 1.7

O:C = 0.27

H:C = 1.2

O:C = 0.09

What about fireworks?

• Many metals were found

associated with

fireworks, with Sr (red

colorant) was found to

be most specific

• Was not part of any

factors, even when

forced by artificial

upweighting, suggesting

fireworks did not

influence the rest of the

data

50

40

30

20

10

0

S

ign

a

l

s

-1

29/10/2014

02/11/2014

06/11/2014

10/11/2014

Date and Time

Sr

Ti

Fe

Cs

Comparisons with other data (Reyes Villegas

et al. 2018)

Bonfire

Factors

Bonfire Factors

Bonfire and Non-Bonfire Factors

00:00

05/11/2014

12:00

00:00

06/11/2014

12:00

Date and Time (UTC)

A

rb

itr

a

ry Un

it (

A

U)

A

rb

itr

a

ry Un

it (

A

U)

rBC

BC

BrC

HONO

HCN

HULIS

Fullerenes

HCNO

HOA

sPON

pPON

Sr

BBOA

Budgeting BC

• The contribution

of the factors to

the overall signal

and the rBC (C

_x

)

signals can be

calculated

• This can be

applied to another

BC measure (e.g.

AE31) to

apportion the BC

Bonfire

factors

Non-Bonfire

factors

1.0

0.8

0.6

0.4

0.2

0.0

Profile Factors

rBC signal fraction

Hydrocarbon-Like Fullerene (42%)

Hydrocarbon-like OA (27%)

Biomass Burning OA (10%)

Domestic Burning (7%)

HULIS (13%)

Residual (-1%)

HO-like Fullerene

HOA

BBOA

Domestic Burning

HULIS

Residual

1.0

0.8

0.6

0.4

0.2

0.0

Profile Factors

Total signal fraction

Hydrocarbon-like Fullerene (11%)

Hydrocarbon-like OA (13%)

Biomass Burning OA (54%)

Domestic Burning (8%)

HULIS (9%)

Residual (5%)

Bonfire

factors

Non-Bonfire

factors

11a

11b

Conclusions

• This case showed that we can use the SP-AMS, incorporating

fullerene signals, to resolve more than two BC sources and budget for

BC

• It is not vital to quantitatively calibrate the SP-AMS providing a

collocated measure of BC is present (e.g. Aethalometer)

• The SP-AMS can also detect firework signals in the form of metals,

with strontium proving the best tracer