In blongworth/HybridGIS: Data, Functions, and Analysis for the NOSAMS Hybrid Gas Ion Source
# Load libraries and functions
library(tidyverse)
library(amstools)
library(here)
library(HybridGIS)
library(broom)
library(patchwork)
theme_set(theme_bw())
options(digits = 3)
Development notes for using the HGIS with an open split. Intention is to
use with gasbench for carbonates and/or large co2 samples.
Capillary parameters
Calculate ideal capillary to provide \~7.5uL/min at atmospheric
pressure.
Short capillary at 1 atm
40um capillary
r <- 2E-5
u <- 1.49E-5
x <- 3.5
flowcalc(100, r, u, x)
r <- 2E-5
u <- 1.49E-5
x <- 2
flowcalc(100, r, u, x)
50um capillary
r <- 2.5E-5
u <- 1.49E-5
x <- 3
flowcalc(100, r, u, x)
r <- 2.5E-5
u <- 1.49E-5
x <- 7
flowcalc(100, r, u, x)
First Tests
2018-04-10 - First tests
Tried a bunch of capillary combinations. Other than initial pressure
burst when opening or closing HGIS valve, couldn't see any source
pressure changes or beam current. We believe the source end of the
stainless capillary was clogged by cesium residue.
Cleaned capillary end at next source cleaning, first with kimwipe and
water, then with isopropanol. Lots of black crud. Confirmed open passage
by blowing N2 through capillary into cup of isopropanol to observe
bubbles.
2018-05-23 - First tests
Tried 3m of 50um glass capillary. Insulated from source to cage with
teflon tubing as outer jacket. No problems with sparks.
Low currents, even when cutting capillary down to 2m. Either steel
capillary is partially clogged or tip is not making good seal with
target.
2018-05-31 - He Dilution
Open split set up with helium dilution. Using 100um glass capillary.
\~50cm source to bulkhead with teflon jacket, 285cm bulkhead to open
split.
Flows
flowcalc(100, r = 2.5E-5, u = 1.49E-5, l = 3.35)
This is way too high. I may have been using a 50um capillary for the
tests.
Pressures
| Condition | Torr |
|--------------|-------|
| base w/cryo | 1E-7 |
| base w/turbo | 4E-7 |
| capillary | 36E-7 |
: Source pressures
Dilution and current
Flow and current tests with varying capillary and dilution. Flows in
mL/min, pressure in Torr, current in uA. Replumbed helium to panel open
split with needle valve partway through to better control flow. Live CO2
set to 10mL/min.
osdil <- read_csv(here("data/2018-05-31_dillution.csv")) %>%
mutate(ratio = helium/co2)
osdil
osdil %>%
filter(caplen == 322) %>%
ggplot(aes(ratio, current)) +
geom_line() +
ggtitle("Current vs dillution ratio for 10 and 18 mL/m CO2") +
facet_grid(co2 ~ .)
Test Acquisitions
Ran 4 test Acquisitions with open split: pure CO2, pure dead CO2,
diluted live, diluted dead. Data are 14/12 ratio with a 13C correction
applied. NormFm is values normalized to the ratio of the solid OX-I.
There are also dual inlet samples on this wheel, filtered out by date.
Positions are incorrect in file because wheel was advanced manually.
Correct positions in notebook and fixed below. Added correct dilution
factors below.
data <- get_hgis_data(here("data/USAMS053018R.txt"),
as.Date(c("2018-05-30", "2018-05-31"))) %>%
filter(Pos %in% c(3,4) | as.Date(ts) == "2018-05-31") %>%
mutate(Pos = case_when(Sample.Name == "LiveGas" ~ 7,
Sample.Name == "DeadGas" ~ 8,
Sample.Name == "LiveGasHe" ~ 9,
Sample.Name == "DeadGasHe" ~ 10,
TRUE ~ as.numeric(Pos)),
dil_factor = case_when(Pos == 9 ~ (50.3 - 9.4)/9.4,
Pos == 10 ~ (40 - 4.8)/4.8),
Num = ifelse(Pos %in% c(3, 4), "S", Num))
# Get mean ratio for solid OX-I and normalize data
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data <- mutate(data, normFm = norm_gas(cor1412he, stdrat))
sum_hgis(data)
plot_hgis(data)
2019-05-31 - startup tests
Reinstalled gas delivery arm after cleaning and polishing ball end.
Using 50um x .5m capillary source to cage and 50um x 2.5m to open split.
Open split flowing modern co2 at 10mL/min. Using same targets and wheel
as last time. Surprisingly good performance, including no-gas base
current, hgis current and solid samples.
flowcalc(pres = 100, r = 2.5E-5, u = 1.49E-5, l = 3.0)
Currents up to 15uA, source vac 8.3E-7 with gas on, 5.5E-7 with gas off.
No data recorded
2019-06-03 - dillution tests
Spun up inlet turbopump. Looked at dilution and current.
dil <- tibble(
CO2 = c(10, 10, 10, 10, 10),
Tot = c(10, 20, 30, 43.5, 10),
vac = c(1.24E-6, 9.52E-7, 8.3E-7, 8E-7, 1.07E-6),
cur = c(10, 9.6, 7, 5, 10)
) %>%
mutate(He = Tot - CO2,
ratio = He/CO2)
dil
ggplot(dil, aes(ratio, cur)) +
geom_line() +
ggtitle("Current vs dillution ratio, CO2 at 10ml/min")
Parameterization of source and open split
2020-09-10 Starting up HGIS
Solid run on pos 1:4 on USAMS012220.
2020-09-29 Dilution vs Ratio
Testing effect of helium dilution on ratio. Used remaining targets on
USAMS012220. Reterminated capillaries at cage wall and at open split.
Used positions 7-10, results in USAMS012220. Low currents: \~6uA pure
CO2, 3uA diluted. Flows: He 10, deadCO2 11, liveCO2 10.
# Load data
data <- get_hgis_data(here("data/USAMS012220R.txt"),
as.Date(c("2020-09-10", "2020-09-29"))) %>%
mutate(dil_factor = ifelse(str_starts(Sample.Name, "Dil"), 1, 0),
Num = ifelse(Pos == 7, "U", Num))
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data <- mutate(data, normFm = norm_gas(cor1412he, stdrat))
sum_hgis(data)
plot_hgis(data)
Looks like diluted and undiluted live gas have the same ratio.
2020-10-15 New wheel, low currents
Loaded a new wheel (USAMS101320) with 4 solid targets and 6 gas targets.
Currents were low (3uA gas, 40uA solid) and didn't increase with tune,
target position or gas arm position. Will clean source and try again.
2020-11-03 Continued dillution tests
Still very low currents, \~2uA for live CO2. Ran a dillution test anyway
with positions 5-8 on USAMS101320.
Flows
| Gas | Flow |
|------|------|
| live | 10 |
| dead | 11.1 |
| He | 30.2 |
data <- get_hgis_data(here("data/USAMS101320R.txt"), as.Date("2020-11-03"))
sum_hgis(data)
plot_hgis(data)
2020-11-17 Dillution testing - back to good currents
Finally back to testing after figuring out current issue. Rerunning
dilution test. Flows: live 11.6mL/min, dead 11.1, He 30.2. Pressures:
cryo base 8E-8, turbo base 3E-7, w/live CO2 1.2E-6. Source producing
\~8uA HE12C with live CO2. Data in USAMS101320.
Running positions 9 and 10 as 1:1 He dilutions of live and dead,
respectively.
data <- get_hgis_data(here("data/USAMS101320R.txt"), as.Date("2020-11-17"))
sum_hgis(data)
Look at counts per second to think about needed acquisition time.
data %>% group_by(Sample.Name) %>%
summarize(cps = mean(CntTotGT/(Cycles/10)))
plot_hgis(data)
data %>% filter(Pos %in% 5:10) %>%
ggplot(aes(he12C, normFm, color = Sample.Name)) +
geom_point()
Doesn't seem to follow curves expected for a MBC with nearly live blank
at about 1% of gas current.
2020-12-04 Blank tests
Ran a series of blank tests before exposing graphite samples or
introducing live co2.
Capillary break between Target blank and Helium blank. Replaced with SGE
100um x 52cm. Initial currents of targets are high, between 10-25uA.
Drops asymptotically to base target current in 5-10 minutes.
data <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2020-12-04"))
sum_hgis(data)
plot_hgis(data)
Modeling Blank
With these data we can start to think about modeling a mass balance, or
more accurately, a current balance blank. It's not initially clear
whether the blank comes from the target/source, diluent helium, or a
combination of the two.
data %>% filter(Pos %in% 9:12) %>%
ggplot(aes(he12C, normFm, color = Sample.Name)) +
geom_point() +
ggtitle("Blank Fm vs. current")
If this blank follows the constant contamination model, plotting Fm vs.
inverse current should give a linear fit. Need more data here to confirm
fit and determine current and Fm of blank.
# subset data
gas <- data %>% filter(Pos %in% 9:12) %>% mutate(icur = 1/he12C)
# fit a linear model
fitlm <- lm(normFm ~ icur, data = gas)
# Add predicted values to data
gas$Fm = predict(fitlm)
ggplot(gas) +
geom_line(aes(1/he12C, y = Fm)) +
geom_point(aes(1/he12C, normFm, color = Sample.Name)) +
ggtitle("Blank Fm vs. inverse current")
TODO: Fit mass balance curve using currents TODO: Use bottle gas to
produce live/dead curves TODO: produce live/dead curves using helium
dilution
2020-12-11 open split flow reduction
Testing reduced open split flows. Worked with flow controller (helium)
and adjusted stops on live and dead co2 needle valves to allow lower
flows. Tested for breakthrough by flowing helium through split and into
source while blowing dead CO2 across the opening of the split at
\~10mL/min. CO2 seen in source somewhere between 100 and 300 uL/min.
Hard to measure with this setup due to slow response time and lack of
sensitivity of capillary/ion source. Need to try this with leak checker.
Reran blank tests at lower flows. Results seem as good or better than
last week's tests at high flows. This may also be due to targets being
more pumped down or burned in.
TODO: get flows from notes.
data <- get_hgis_data(here("data/USAMS120320R.txt")) %>%
filter(Pos %in% 1:4 | as.Date(ts) == "2020-12-11") %>%
mutate(he = case_when(Pos == 9 ~ (1059 - 383),
Pos == 10 ~ 0,
Pos == 11 ~ 1370,
Pos == 12 ~ 0),
co2 = case_when(Pos == 9 ~ 383,
Pos == 10 ~ 383,
Pos == 11 ~ 0,
Pos == 12 ~ 0),
dil_rat = he/co2)
data %>%
group_by(Pos, Sample.Name, dil_rat) %>%
summarise(Cur = mean(he12C),
Cur.sd = sd(he12C),
mean = mean(normFm),
sd = sd(normFm),
interr = 1/sqrt(sum(CntTotGT)),
acqtime = sum(Cycles)/10,
N_acq = n())
plot_hgis(data)
# subset data
gas <- data %>% filter(Pos %in% 9:12) %>% mutate(icur = 1/he12C)
# fit a linear model
fitlm <- lm(normFm ~ icur, data = gas)
# Add predicted values to data
gas$Fm = predict(fitlm)
ggplot(gas) +
geom_line(aes(1/he12C, Fm)) +
geom_point(aes(1/he12C, normFm, color = Sample.Name)) +
ggtitle("Blank Fm vs. inverse current")
2020-12-18 More testing of low open split flow rates
Tested open split breakthrough at low flows using leak checker. Measured
helium flow through capillary using leak checker. Ran target blanks,
dead and live samples at low flows. Started building double needle
testbed.
Leak checker flow rate
Connected cage-split capillary to leak checker. Base leak rate with
capillary in air: 1.7E-9 mbarLs-1. Leak rate with helium in open split:
3.5E-4 mbarls-1.
Leak checker breakthrough tests
Flowing dead CO2 through split. Flowing helium across mouth of split at
8mLmin-1. Detection of helium means atmosphere is getting into split.
lc_breakthrough <- tribble(~co2, ~leakrate,
730, 6E-9,
735, 6.5E-9,
525, 7.9E-9,
178, 1.3E-7,
150, 2E-7,
0, 3E-6)
lc_breakthrough
ggplot(lc_breakthrough, aes(co2, leakrate)) + geom_line() + geom_point() +
ggtitle("Helium leak rate vs. CO2 flow to split")
Data acquisitions at low(er) flows
Source off base: 4E-8 Cryo base (HV + Ion): 6E-8 Turbo base: 2E-7
Set SRS to 10uA/V and 50nA/V for 12 and 13C.
Flows:
flows <- tribble(~pos, ~Sample.Name, ~co2, ~he,
10, "DeadGas", 270, 0,
9, "DilDeadGas", 112, 288,
8, "DilDeadGas6", 102, 0,
8, "DeadGas", 102, 572-102,
7, "HeliumBlank", 0, 572-102,
7, "DilLiveGas6", 117, 576-117) %>%
mutate(dil_ratio = he/co2)
flows
Data
data <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2020-12-18"))
sum_hgis(data)
plot_hgis(data)
data %>% filter(Pos %in% 7:10) %>%
ggplot(aes(1/he12C, normFm, color = Sample.Name)) +
geom_point() +
ggtitle("Fm of live and dead samples vs. inverse current")
2021-01-08 Pure gas precison testing
Plumbed long 50um capillary from live CO2 regulator to cage wall
fitting. This allows controlling flow of CO2 to source using regulator
pressure.
Current vs flow and pressure
First test was with 748cm x 50 um capillary. Source base pressure using
the cryopump was 2E-7 Torr. Base current for Pos 7 was 300uA. Initial
max current \@ 200kPA CO2 was 9uA. Did not return to this level in the
pressure test following, not sure why, source pressure shows CO2 was
getting to source. Typically best currents with pure CO2 are at a source
pressure of around 1E-6 Torr. Reg pressure in kPa, source in Torr x 10-7
and current in uA.
cur_vs_pres <- tribble(~regp, ~sourcep, ~c12cur,
0, 2.8, 0.8,
50, 3.3, 1,
100, 5.3, 1.2,
150, 8.2, 1.4,
200, 12, 1.8,
250, 17, 2.4,
300, 22, 3.1,
350, 28, 3.7,
400, 35, 4.5)
cur_vs_pres
Tested again with new capillaries after a clog in regulator-cage
capillary and spark damage to cage-source capillary. New capillaries:
reg-cage- 639cm x 50um, cage-source- 50cm x 50um. Base current 250nA.
cur_vs_pres2 <- tribble(~regp, ~sourcep, ~c12cur,
25, 2, .25,
100, 4, 1.3,
150, 8.4, 6.3,
200, 12, 6.2,
250, 17, 6.2)
cur_vs_pres2
Data vs flow
Took data on targets 6 and 7 at various CO2 flow rates.
Data summary
data <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2021-01-08"))
ds <- sum_hgis(data)
ds
plot_hgis(data)
Ratio and current vs flow
Took data for 4 different flows on target 6. CO2 flow data show ratio
dependence on flow. Need to fill in data to get shape of curve. Current
vs flow
ds %>%
filter(Pos == 6) %>%
mutate(Pressure = as.numeric(str_extract(Sample.Name, "\\d+"))) %>%
ggplot(aes(Pressure, Cur)) +
geom_pointrange(aes(ymin = Cur - Cur.sd, ymax = Cur + Cur.sd)) +
geom_point(size = 4) +
ggtitle("Fm of live CO2 vs. C12 current")
Fm vs. Flow. Note that solid standards had a lot of variability, so
treat ratio as relative, not well normalized.
ds %>%
filter(Pos == 6) %>%
mutate(Pressure = str_extract(Sample.Name, "\\d+")) %>%
ggplot(aes(Cur, mean, shape = Pressure, color = Pressure)) +
geom_pointrange(aes(ymin = mean - sd, ymax = mean + sd)) +
geom_point(size = 4) +
ggtitle("Fm vs. C12 current for varying CO2 regulator pressure")
2021-01-15 Precision and vial tests
Used test vial setup to test pure CO2 precision and introduction of CO2
from vial by displacing with helium.
Data
df <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2021-01-15"))
stdrat <- mean(df$cor1412he[df$pos_name == "8 - LiveGasOS" & !df$outlier])
df <- mutate(df, normFm = norm_gas(cor1412he, stdrat))
sum_hgis(df)
Ratio and current during runs
plot_hgis_time(df) +
facet_wrap(vars(pos_name), scales = "fixed")
ggplot(df, aes(cum_acqtime, he12C )) +
geom_point() +
facet_wrap(vars(pos_name), scales = "fixed")
Ratio dependence on current
ggplot(df, aes(he12C, normFm, color = pos_name)) +
geom_point() +
ggtitle("Fm of modern CO2 vs. C12 current")
ggplot(df, aes(1/he12C, normFm, color = pos_name)) +
geom_point() +
ggtitle("Fm of modern CO2 vs. inverse C12 current")
Question: Is ratio current depletion change in CO2 from vial or target
degradation? Long pure CO2 run to test.
2021-01-22 Pure gas precison testing
Plumbed long 50um capillary from Valco valve to cage wall fitting.
Plumbed dead CO2 to Valco with 1/16" ss.
Took data with dead CO2 on target 5 at various CO2 flow rates. Same with
liveCO2 on target 6. Source not performing well, low currents and
current did not correspond well to CO2 pressure/flow. Current vs ratio
data may still be useful for blank model. Solid currents good, so this
is a gas issue. Gas blank also not great.
Data summary
data <- get_hgis_data(here("data/USAMS012121R.txt"), as.Date("2021-01-22"))
ds <- sum_hgis(data)
ds
Note that solid standards had a lot of variability, so treat ratio as
relative, not well normalized.
plot_hgis(data)
Relationship of current and ratio to CO2 flow rate
Current vs flow
Supply pressure is proportional to flow rate.
ds <- ds %>%
filter(Pos %in% c(5, 6)) %>%
mutate(Cur_inv = 1/Cur,
Pressure = as.numeric(str_extract(Sample.Name, "\\d+")),
Gas = ifelse(Pos == 5, "DeadCO2", "LiveCO2"))
ggplot(ds, aes(Pressure, Cur, color = Gas)) +
geom_pointrange(aes(ymin = Cur - Cur.sd, ymax = Cur + Cur.sd)) +
geom_point(size = 4) +
ggtitle("C12 current vs. CO2 regulator pressure")
Fm vs. current and flow rate
ggplot(ds, aes(Cur, mean, color = Pos)) +
geom_pointrange(aes(shape = factor(Pressure), ymin = mean - sd, ymax = mean + sd)) +
ggtitle("Fm of live and dead gas vs C12 current")
Modelling current dependent blank
Fit linear models
fits <- ds %>%
ungroup() %>%
nest(data = -Gas) %>%
mutate(fit = map(data, ~lm(mean ~ Cur_inv, data = .x)),
tidied = map(fit, tidy)) %>%
unnest(tidied)
fits %>%
select(Gas, term, estimate) %>%
pivot_wider(names_from = c(Gas, term), values_from = estimate) %>%
mutate(inv_m_blank = -(`DeadCO2_(Intercept)` - `LiveCO2_(Intercept)`)/(DeadCO2_Cur_inv - LiveCO2_Cur_inv),
Fm_blank = `DeadCO2_(Intercept)` + DeadCO2_Cur_inv * inv_m_blank,
m_blank = 1/inv_m_blank) %>%
select(Fm_blank, m_blank)
Fm vs. inverse mass with fits.
ggplot(ds, aes(Cur_inv, mean, color = Pos)) +
geom_pointrange(aes(shape = factor(Pressure), ymin = mean - sd, ymax = mean + sd)) +
geom_smooth(method = "lm", fullrange = TRUE) +
xlim(0, 5) +
ggtitle("Fm of live and dead CO2 vs. inverse 12C current",
subtitle = "Fit lines should cross at Fm and current of blank")
For these measurements, it looks like the blank is roughly 225nA at half
modern. This is higher current and lower Fm than expected based on
previous measurements. This is in line with the high blanks seen when
measuring dead carbon.
Gas delivery from vials
2021-01-29 Reproducibility between multiple vials
Measured live and dead CO2 from vials. "STD" samples were displaced to
split with flowing CO2, while the "V" samples were displaced by helium
as they would be for unknowns. All samples measured on fresh targets
after a 5 minute "burn in" to reduce blank.
Gas flows
Replaced capillary from Valco valve to double needle with longer piece
of 50um capillary. Set helium and CO2 flows.
+----------------------+-----------------------+-----------------------+
| Gas | Pressure (kPa) | Flow (uL/min) |
+======================+=======================+=======================+
| LiveCO2 | 165 | 136 |
+----------------------+-----------------------+-----------------------+
| DeadCO2 | 160 | 120 |
+----------------------+-----------------------+-----------------------+
| Helium | 165 | 115 |
+----------------------+-----------------------+-----------------------+
: Gas flow rates
Measured split to cage wall capillary flow with leak checker. Leak rate:
1.0E-5 Pa m\^3/s.
Data
Normalizing to solid OX-I's.
data <- get_hgis_data(here("data/USAMS012121R.txt"), as.Date("2021-01-29")) %>%
mutate(Num = ifelse(Pos %in% c(7, 9), "U", Num))
Currents and ratios vs time.
Sample ratio vs. time
plot_hgis_time(data, cor1412he, ce*cor1412he)
Sample current vs time
plot_hgis_time(data, he12C)
# Get mean ratio for solid OX-I and normalize data
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data <- mutate(data, normFm = norm_gas(cor1412he, stdrat))
Per-sample summary
ds <- data %>%
filter(!outlier) %>%
sum_hgis()
ds
plot_hgis(data)
Agreement of replicates
Summary statistics to look at reproducibility between targets/vials.
ds %>%
filter(as.numeric(Pos) > 4,
Pos != 9) %>% # Don't use bad target
mutate(gas = ifelse(str_starts(Sample.Name, "Live"), "Live", "Dead"),
helium = ifelse(str_detect(Sample.Name, "V\\d$"), TRUE, FALSE)) %>%
group_by(gas, helium) %>%
summarize(Fm.mean = mean(mean),
Fm.sd = sd(mean),
N = n())
It looks like the SD of modern samples is around 1%, and that pure CO2
performs slightly better than helium displacement.
2021-02-05 Carbonate samples
Measured carbonate samples prepared on gasbench. Samples included C-1,
TIRI-F, TIRI-I, C-2, and NOSAMS2. Normalized to gas standards or solid
OX-I.
First carbonate sample clogged capillary. Capillary replaced, moved
needle up to avoid sucking in acid. Was not able to get good currents
after replacing capillary, even trying several length/capillary
combinations. Source vacuum indicates normal amount of gas entering
source. May need to consider a water trap. All samples measured on fresh
targets after a 5 minute "burn in" to reduce blank.
Samples
carb_data <- tribble(~vial, ~name, ~rec_num, ~mass,
1, "C-1", 83028, 32.05,
2, "TIRI-F", 2138, 33.56,
3, "TIRI-I", 17185, 32.67,
4, "TIRI-I", 17185, 32.35,
5, "C-2", 1082, 31.35,
6, "C-2", 1082, 34.11,
7, "NOSAMS2", 38809, 32.30,
8, "NOSAMS2", 38809, 34.00
)
write_csv(carb_data, here("data/carb_data_2021-02-05.csv"))
carb_data
Gas flows
Set helium and CO2 flows.
+----------------------+-----------------------+-----------------------+
| Gas | Pressure (kPa) | Flow (uL/min) |
+======================+=======================+=======================+
| LiveCO2 | 150 | 95 |
+----------------------+-----------------------+-----------------------+
| Helium | 150 | 97 |
+----------------------+-----------------------+-----------------------+
: Gas flow rates
Data
Normalizing to gas standards in pos 5 and 8.
data <- get_hgis_data(here("data/USAMS020521R.txt")) %>%
mutate(Num = ifelse(Pos %in% c(2, 4), "U", Num))
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data <- data %>%
mutate(normFm = norm_gas(cor1412he, stdrat))
Currents and ratios vs time.
Sample ratio vs. time
plot_hgis_time(data, normFm, ce*normFm)
Sample current vs time
plot_hgis_time(data, he12C)
Summary of normalized data per sample
ds <- data %>%
filter(!outlier) %>%
sum_hgis() %>%
mutate(rec_num = case_when(str_starts(Sample.Name, "LiveGas") ~ 101730,
Sample.Name == "C-1" ~ 83028,
Sample.Name == "TIRI-F" ~ 2138,
Sample.Name == "TIRI-I" ~ 17185,
Sample.Name == "C-2" ~ 1082,
Sample.Name == "NOSAMS2" ~ 38809))
std <- getStdTable()
ds <- left_join(ds, select(std, rec_num, fm_consensus)) %>%
mutate(fm_consensus = ifelse(rec_num == 101730, 1.0398, fm_consensus))
ds[-(1:4),-3]
data %>%
filter(as.numeric(Pos) > 4) %>%
ggplot(aes(pos_name, normFm)) +
geom_boxplot(fill = "slategray1") +
theme(axis.text.x = element_text(angle = 45, hjust=1))
Agreement of replicates
Summary statistics to look at reproducibility, accuracy between
targets/vials.
ds %>%
filter(as.numeric(Pos) > 4,
Pos != 6) %>% # Don't use bad target
mutate(Name = ifelse(Pos %in% c(5,8), "LiveGas", Sample.Name)) %>%
group_by(Name, fm_consensus) %>%
summarize(Fm.mean = mean(mean),
Fm.sd = sd(mean),
Cur.mean = mean(Cur),
N = n())
It looks like the SD of replicates of modern samples with current > 3uA
is around 0.5%, which is better than measurement statistics. Fm of
NOSAMS 2 is a bit high. Need to investigate current-dependent blank.
Blank correction
Testing correcting for blank using a simple, mass-independent large
blank correction.
LBC does not work well for this set of samples.
fmblank <- ds$mean[ds$Sample.Name == "TIRI-F"]
fmblank
ds <- ds %>%
mutate(fm_lbc = doLBC(mean, fmblank, 1.0398))
ds %>%
filter(as.numeric(Pos) > 4) %>%
select(Sample.Name, Cur, fm_consensus, mean, fm_lbc)
ds %>%
write_csv(here("data_analysed/USAMS020521.csv"))
2021-03-01 Live gas and solid samples
Ran live CO2 from vial to check performance after source cleaning and
cryopump issues. Slow to get going, so only one modern gas standard and
solid samples run. Tried second modern gas standard on pos 16, but no
current from gas that worked on pos 15. Bad target?
data <- get_hgis_data(here("data/USAMS020521R.txt"), as.Date("2021-03-01"))
Currents and ratios vs time.
Sample ratio vs. time
plot_hgis_time(data, normFm, ce*normFm)
Sample current vs time
plot_hgis_time(data, he12C)
Normalization
# Get mean ratio for solid OX-I and normalize data
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data <- data %>%
mutate(normFm = norm_gas(cor1412he, stdrat))
Summary of normalized data per sample
ds <- data %>%
filter(!outlier) %>%
sum_hgis() %>%
mutate(rec_num = case_when(str_starts(Sample.Name, "LiveGas") ~ 101730,
Sample.Name == "C-1" ~ 83028,
Sample.Name == "TIRI-F" ~ 2138,
Sample.Name == "TIRI-I" ~ 17185,
Sample.Name == "C-2" ~ 1082,
Sample.Name == "NOSAMS2" ~ 38809))
std <- getStdTable()
ds <- left_join(ds, select(std, rec_num, fm_consensus)) %>%
mutate(fm_consensus = ifelse(rec_num == 101730, 1.0398, fm_consensus))
ds
plot_hgis(data)
2021-03-05 Carbonate samples
Measured carbonate samples prepared on gasbench. Samples included C-1,
C-2, and NOSAMS2. Normalized to gas standards or solid OX-I.
All samples measured on fresh targets after a 5 minute "burn in" to
reduce blank. Gas flows same as 2021-02-05.
Samples
carb_data <- tribble(~vial, ~name, ~rec_num, ~mass,
1, "C-1", 83028, 33.69,
2, "C-1", 83028, 31.62,
3, "C-2", 1082, 32.70,
4, "C-2", 1082, 32.08,
5, "NOSAMS2", 38809, 33.23,
6, "NOSAMS2", 38809, 33.95
)
write_csv(carb_data, here("data/carb_data_2021-03-05.csv"))
carb_data
Data
Normalizing to gas standards in pos 5 and 8.
data <- get_hgis_data(here("data/USAMS030421R.txt"), as.Date("2021-03-05"))
Currents and ratios vs time.
Sample ratio vs. time
plot_hgis_time(data, normFm, ce*normFm)
Sample current vs time
plot_hgis_time(data, he12C)
Summary of normalized data per sample
std <- getStdTable()
ds <- data %>%
filter(!outlier) %>%
sum_hgis() %>%
mutate(rec_num = case_when(str_starts(Sample.Name, "LiveGas") ~ 101730,
str_starts(Sample.Name, "C1") ~ 83028,
str_starts(Sample.Name, "TIRI-F") ~ 2138,
str_starts(Sample.Name, "TIRI-I") ~ 17185,
str_starts(Sample.Name, "C2") ~ 1082,
str_starts(Sample.Name, "NOSAMS2") ~ 38809,
str_starts(Sample.Name, "DeadGas") ~ 72446)) %>%
left_join(select(std, rec_num, fm_consensus), by = "rec_num") %>%
mutate(fm_consensus = case_when(rec_num == 101730 ~ 1.0398,
rec_num == 72446 ~ 0.0013,
TRUE ~ fm_consensus))
ds[-(1:4),-3]
data %>%
filter(as.numeric(Pos) > 4) %>%
ggplot(aes(pos_name, normFm)) +
geom_boxplot(fill = "slategray1") +
theme(axis.text.x = element_text(angle = 45, hjust=1))
Agreement of replicates
Summary statistics to look at reproducibility, accuracy between
targets/vials.
Mean of modern CO2 filled on HGIS setup (LiveGas) and Gilson
(LiveGasGil) were used for normalization but were fairly different on
this run, so shown separately.
ds %>%
filter(as.numeric(Pos) > 4) %>%
mutate(Name = str_remove(Sample.Name, "-.$")) %>%
group_by(Name, fm_consensus) %>%
summarize(Fm.mean = mean(mean),
Fm.sd = sd(mean),
Cur.mean = mean(Cur),
N = n())
Agreement of replicates is better than 2% for all samples. Samples have
higher Fm than consensus. I think most of this is due to blanks, but the
low gas standard also would push unknown Fm's up.
Blank correction
Testing correcting for blank using a simple, mass-independent large
blank correction.
Using dead gas blank, naive with no adjustment for graphite value of
blank. This pushes Fm in the right direction, but comparing samples of
the same type, there's a lot of variability that seems to be tied to
sample current.
fmblank <- ds$mean[ds$Sample.Name == "DeadGas"]
fmblank
ds <- ds %>%
mutate(fm_lbc = doLBC(mean, fmblank, 1.0398))
ds %>%
filter(as.numeric(Pos) > 4) %>%
select(Sample.Name, Cur, fm_consensus, mean, fm_lbc)
ds %>%
write_csv(here("data_analysed/USAMS030421.csv"))
2021-03-26 High dillution test
Determine length of 75um capillary for \~100ulm to source.
r <- 7.5E-5 / 2
u <- 1.49E-5
x <- 3
flowcalc(100, r, u, x)
Startup and test after opening tank to refill terminal argon bottle.
Testing helium dilution by starting with helium in vial and displacing
with CO2. Used 2.5m of 75um capillary to provide 100ulm to source, but
initial source pressure of 3E-7 Torr indicates lower flows. Shortened
split-cage capillary to about 150cm, without a huge effect on pressure.
Flow may be defined by cage-source capillary.
LiveCO2 flow set to 244 uL/m.
Using position 14 on USAMS030421. Initial current after 10min: 150nA.
Currents rose to a stable 10uA and stayed there for more than 90 min!
Data
data <- get_hgis_data(here("data/USAMS030421R.txt"), as.Date("2021-03-26")) %>%
mutate(Num = ifelse(Pos %in% c(2,4), "S", Pos))
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data <- data %>%
mutate(normFm = norm_gas(cor1412he, stdrat))
Currents and ratios vs time.
data %>%
filter(Pos == 14) %>%
plot_hgis_time(normFm, ce*normFm) +
ggtitle("Sample ratio vs. time")
data %>%
filter(Pos == 14) %>%
plot_hgis_time(he12C) +
ggtitle("Sample current vs. time")
Predicted flow of CO2 to source for test flow conditions
co2flow <- 244 / 60 #flow in ul/s
data.frame(x = 0:8000) %>%
ggplot(aes(x)) +
stat_function(fun=function(x) concCO2(x, r = co2flow) * 7, aes(color = "He")) +
stat_function(fun=function(x) 7 - concCO2(x, r = co2flow) * 7, aes(color = "CO2")) +
scale_color_manual("Gas", values = c("green", "blue")) +
labs(title = "CO2 in vial over time",
subtitle = "7mL vial, 244uL/min CO2",
x = "Time (s)",
y = "Volume (mL)")
Stable currents reached at around 5mL CO2 in vial, but stable ratios
reached at around 2mL. Contrast with helium dilution test from 4/2
below.
Ratio vs Current
data %>%
filter(Pos == 14) %>%
ggplot(aes(he12C, normFm)) +
geom_point()
2021-04-02 High dillution test
Testing performance at higher dilution ratio. Started with pure live CO2
in vial and did long acquisition with high displacement helium flow to
check current as CO2 concentration decreases.
Helium flow set to 244 uL/m.
Using position 25 on USAMS040321. Shifted targets to higher positions to
avoid wearing only low wheel positions. Initial current after 10min:
150nA.
Data
data <- get_hgis_data(here("data/USAMS040121R.txt"), as.Date("2021-04-02")) %>%
mutate(Num = ifelse(Pos == 5, "S", Pos))
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data <- data %>%
mutate(normFm = norm_gas(cor1412he, stdrat))
Currents and ratios vs time.
Sample ratio vs. time
data %>%
filter(Pos == 5) %>%
plot_hgis_time(normFm, ce*normFm)
Sample current vs time
data %>%
filter(Pos == 5) %>%
plot_hgis_time(he12C)
Expected CO2 ratio/flow
Plot volume of CO2 in sample vial during displacement by helium. It
looks like optimal currents are maintained until about 2000s, which
corresponds to about 2mL of CO2. Stable ratios until about 1.25 mL CO2.
Can we back calculate flow to source if current falls off below
\~2ul/min CO2?
heflow <- 244 / 60 #flow in ul/s
data.frame(x = 0:8000) %>%
ggplot(aes(x)) +
stat_function(fun=function(x) concCO2(x, r = heflow) * 7, aes(color = "CO2")) +
scale_color_manual("Gas", values = c("green", "blue")) +
labs(title = "CO2 in vial over time",
subtitle = "7mL vial, 244uL/min helium",
x = "Time (s)",
y = "CO2 (mL)")
data.frame(x = 0:8000) %>%
ggplot(aes(x)) +
stat_function(fun=function(x) concCO2(x, r = heflow) * 30, aes(color = "CO2")) +
scale_color_manual("Gas", values = c("green", "blue")) +
labs(title = "CO2 flow over time",
subtitle = "7mL vial, 244uL/min helium",
x = "Time (s)",
y = "CO2 (uL/s)")
Add flow at time to data and plot current vs flow. 30 ul/m is estimate
of capillary flow based on past results.
data <- data %>%
mutate(co2flow = concCO2(cum_acqtime, r = heflow) * 30) #30ul/min
ggplot(data, aes(co2flow, he12C)) +
geom_smooth(span = .3, se=FALSE) +
geom_point() +
labs(title = "Vial dillution current",
subtitle = "250uL/min displacement, 30ul/min delivery",
x = "CO2 flow (ul/m)",
y = "12C current (uA)")
Ratio vs Current
data %>%
filter(Pos == 5) %>%
ggplot(aes(he12C, normFm)) +
geom_point()
2021-04-09 Carbonate samples
Measured carbonate samples prepared on gasbench. Samples included
TIRI-F, TIRI-I, and NOSAMS2. Normalized to gas standards or solid OX-I.
All samples measured on fresh targets after a 10 minute "burn in" to
reduce blanks.
Using short capillaries from 2021-04-02. He flow set to 250ul/min.
Samples
carb_data <- tribble(~vial, ~name, ~rec_num, ~mass,
1, "NOSAMS2_1", 38809, 31.8,
2, "NOSAMS2_2", 38809, 32.3,
3, "NOSAMS2_3", 38809, 33.5,
4, "TIRI-I_1", 17185, 33.1,
5, "TIRI-I_2", 17185, 31.8,
6, "TIRI-I_3", 17185, 31.8,
7, "TIRI-F_1", 2138, 31.9,
8, "TIRI-F_2", 2138, 33.4,
9, "TIRI-F_3", 2138, 32.4
)
write_csv(carb_data, here("data/carb_data_2021-04-08.csv"))
carb_data
Data
Normalizing to gas standards in pos 26 and 37.
data <- get_hgis_data(here("data/USAMS040121R.txt"), as.Date("2021-04-09")) %>%
mutate(Num = ifelse(Pos %in% c(22, 24, 25, 27, 38), "U", Num))
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data <- data %>%
mutate(normFm = norm_gas(cor1412he, stdrat))
Currents and ratios vs time.
plot_hgis_time(data, normFm, ce*normFm) +
ggtitle("Sample ratio vs. time")
plot_hgis_time(data, he12C) +
ggtitle("Sample current vs. time")
Summary of normalized data per sample
std <- getStdTable()
ds <- data %>%
filter(!outlier) %>%
sum_hgis() %>%
mutate(rec_num = case_when(str_starts(Sample.Name, "LiveGas") ~ 101730,
str_starts(Sample.Name, "C1") ~ 83028,
str_starts(Sample.Name, "TIRI-F") ~ 2138,
str_starts(Sample.Name, "TIRI-I") ~ 17185,
str_starts(Sample.Name, "C2") ~ 1082,
str_starts(Sample.Name, "NOSAMS2") ~ 38809,
str_starts(Sample.Name, "DeadGas") ~ 72446)) %>%
left_join(select(std, rec_num, fm_consensus), by = "rec_num") %>%
mutate(fm_consensus = case_when(rec_num == 101730 ~ 1.0398,
rec_num == 72446 ~ 0.0013,
TRUE ~ fm_consensus))
ds[-(1:4),-4]
data %>%
filter(as.numeric(Pos) > 5) %>%
ggplot(aes(pos_name, normFm)) +
geom_boxplot(fill = "slategray1") +
theme(axis.text.x = element_text(angle = 45, hjust=1))
Agreement of replicates
Summary statistics to look at reproducibility, accuracy between
targets/vials.
ds %>%
filter(as.numeric(Pos) > 5) %>%
mutate(Name = str_remove(Sample.Name, "_.$")) %>%
group_by(Name, fm_consensus) %>%
summarize(Fm.mean = mean(mean),
Fm.sd = sd(mean),
Cur.mean = mean(Cur),
N = n())
Agreement of replicates is better than 2% for all samples. Samples have
higher Fm than consensus. I think most of this is due to blanks, but the
low gas standard also would push unknown Fm's up.
Blank correction
Testing correcting for blank using a mass-independent large blank
correction.
Using TIRI-F as blank, naive with no adjustment for graphite value of
blank. This pushes Fm in the right direction, but comparing samples of
the same type, there's a lot of variability that seems to be tied to
sample current.
fmblank <- ds %>%
filter(str_starts(Sample.Name, "TIRI-F")) %>%
pull(mean) %>%
mean()
fmblank
ds <- ds %>%
mutate(fm_lbc = doLBC(mean, fmblank, 1.0398),
diff = fm_lbc - fm_consensus,
normFm = normFm(fm_lbc, fm_consensus),
sigma = sigma(fm_lbc, fm_consensus, merr))
ds %>%
filter(as.numeric(Pos) > 4) %>%
select(Sample.Name, Cur, fm_consensus, mean, fm_lbc, diff)
ds %>%
write_csv(here("data_analysed/USAMS040121.csv"))
ggplot(ds, aes(fm_consensus, diff)) +
geom_hline(yintercept = 0) +
geom_errorbar(aes(ymin = diff - merr, ymax = diff + merr)) +
geom_point() +
ggtitle("Difference in LBC-corrected vs consensus")
ggplot(ds, aes(fm_consensus, sigma)) +
geom_hline(yintercept = 0) +
geom_point() +
ggtitle("Sigma of LBC-corrected vs consensus")
It looks like the normalization may be off a bit. Note positive linear
correlation of Fm consensus difference with Fm of sample. Gas standards
run at end of day were higher than those from the beginning. NOSAMS-2
was run early in the day, so closer to standards with lower ratios.
2021-04-16 Carbonate samples
Measured carbonate samples prepared on gasbench. Samples included C-1,
C-2, and NOSAMS2. Normalized to gas standards or solid OX-I.
All samples measured on fresh targets after a 10 minute "burn in" to
reduce blank.
Using short capillaries from 2021-04-02. He flow set to 250ul/min. Had
to shorten delivery capillary by \~10cm at start of first C-2 due to
acid clog. Needle blown out with compressed air.
Samples
carb_data <- tribble(~vial, ~name, ~rec_num, ~mass,
1, "C-1_1", 83028, 30.05,
2, "C-1_2", 83028, 36.00,
3, "C-1_3", 83028, 31.88,
4, "C-2_1", 1082, 32.57,
5, "C-2_2", 1082, 35.50,
6, "C-2_3", 1082, 34.10,
7, "NOSAMS-2_1", 38809, 35.20,
8, "NOSAMS-2_2", 38809, 32.19,
9, "NOSAMS-2_3", 38809, 33.20
)
write_csv(carb_data, here("data/carb_data_2021-04-15.csv"))
carb_data
Data
Normalizing to gas standards in pos 25, 26, 36 and 37.
data <- get_hgis_data(here("data/USAMS041521R.txt"), as.Date("2021-04-16"))
Currents and ratios vs time.
data %>%
filter(!(Pos %in% 21:24)) %>%
plot_hgis_time(normFm, ce*normFm) +
ggtitle("Sample ratio vs. time")
data %>%
filter(!(Pos %in% 21:24)) %>%
plot_hgis_time(he12C)
ggtitle("Sample current vs time")
Summary of normalized data per sample
std <- getStdTable()
ds <- data %>%
filter(!outlier) %>%
sum_hgis() %>%
mutate(rec_num = case_when(str_starts(Sample.Name, "LiveGas") ~ 101730,
str_starts(Sample.Name, "C-1") ~ 83028,
str_starts(Sample.Name, "TIRI-F") ~ 2138,
str_starts(Sample.Name, "TIRI-I") ~ 17185,
str_starts(Sample.Name, "C-2") ~ 1082,
str_starts(Sample.Name, "NOSAMS-2") ~ 38809,
str_starts(Sample.Name, "DeadGas") ~ 72446)) %>%
left_join(select(std, rec_num, fm_consensus), by = "rec_num") %>%
mutate(fm_consensus = case_when(rec_num == 101730 ~ 1.0398,
rec_num == 72446 ~ 0.0013,
TRUE ~ fm_consensus))
ds[-(1:4),-3]
data %>%
filter(as.numeric(Pos) > 4) %>%
ggplot(aes(pos_name, normFm)) +
geom_boxplot(fill = "slategray1") +
theme(axis.text.x = element_text(angle = 45, hjust=1))
Agreement of replicates
Summary statistics to look at reproducibility, accuracy between
targets/vials.
ds %>%
filter(as.numeric(Pos) > 4) %>%
mutate(Name = str_remove(Sample.Name, "_.$")) %>%
group_by(Name, fm_consensus) %>%
summarize(Fm.mean = mean(mean),
Fm.sd = sd(mean),
Cur.mean = mean(Cur),
N = n())
Blank correction
Testing correcting for blank using mass-independent large blank
correction.
Using C-1 with no adjustment for graphite value of blank.
fmblank <- ds %>%
filter(str_starts(Sample.Name, "C-1")) %>%
pull(mean) %>%
mean()
fmblank
ds <- ds %>%
mutate(fm_lbc = doLBC(mean, fmblank, 1.0398),
diff = fm_lbc - fm_consensus,
normFm = normFm(fm_lbc, fm_consensus),
sigma = sigma(fm_lbc, fm_consensus, merr))
ds %>%
filter(as.numeric(Pos) > 4) %>%
select(Sample.Name, Cur, fm_consensus, mean, merr, fm_lbc, diff)
ds %>%
write_csv(here("data_analysed/USAMS041521.csv"))
ggplot(ds, aes(fm_consensus, diff)) +
geom_hline(yintercept = 0) +
geom_errorbar(aes(ymin = diff - merr, ymax = diff + merr)) +
geom_point() +
ggtitle("Difference in LBC-corrected vs consensus")
ggplot(ds, aes(fm_consensus, sigma)) +
geom_hline(yintercept = 0) +
geom_point() +
ggtitle("Sigma of LBC-corrected vs consensus")
There may be some issues with normalization. The initial two standards
on both of the last carbonate runs were lower than the two at the end of
the run. On 4/9, first two standards were Fm 1.03 and 0.99, last two
were 1.05 and 1.07. On 4/16, first two were 1.02 and 1.04, last two were
both 1.05. NOSAMS2 were run first on 4/9 and were high after blank
correction. On 4/16 they were run last and were low.
Work for SNEAP 2021 presentation
Combining data
A weak attempt at combining data. The solution is an analysis script to
normalize, average runs, and blank correct by wheel.
ADD MOAR DATA
data<- get_hgis_data(here("data/USAMS101320R.txt"), as.Date("2020-11-17")) %>%
mutate(Num = ifelse(Pos %in% c(2, 4), "S", ifelse(Num == "S", "U", Num)))
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data1013 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat))
data <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2020-12-04")) %>%
mutate(Num = ifelse(Pos %in% c(2, 4), "S", ifelse(Num == "S", "U", Num)))
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data1204 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat))
data1211 <- get_hgis_data(here("data/USAMS120320R.txt")) %>%
filter(Pos %in% 1:4 | as.Date(ts) == "2020-12-11")
data <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2020-12-18")) %>%
filter(Pos != 0) %>%
mutate(Num = ifelse(Pos %in% c(2, 4), "S", ifelse(Num == "S", "U", Num)))
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data1218 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat))
data0108 <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2021-01-08"))
# no solid standards
# df <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2021-01-15"))
# stdrat <- mean(df$cor1412he[df$pos_name == "8 - LiveGasOS" & !df$outlier])
# data0115 <- mutate(df, normFm = norm_gas(cor1412he, stdrat))
data <- get_hgis_data(here("data/USAMS020521R.txt")) %>%
mutate(Num = ifelse(Pos %in% c(2, 4), "S", ifelse(Num == "S", "U", Num)))
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data205 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat))
data <- get_hgis_data(here("data/USAMS030421R.txt"), as.Date("2021-03-05")) %>%
mutate(Num = ifelse(Pos %in% c(2, 4), "S", ifelse(Num == "S", "U", Num)))
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data304 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat))
data <- get_hgis_data(here("data/USAMS040121R.txt"), as.Date("2021-04-09")) %>%
mutate(Num = ifelse(Pos %in% c(22, 24), "S", ifelse(Num == "S", "U", Num)))
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data409 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat))
data <- get_hgis_data(here("data/USAMS041521R.txt"), as.Date("2021-04-16")) %>%
mutate(Num = ifelse(Pos %in% c(22, 24), "S", ifelse(Num == "S", "U", Num)))
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data415 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat))
std <- getStdTable()
data <- rbind(data1013, data1204, data1211, data1218, data205, data304, data409, data415) %>%
mutate(rec_num = case_when(str_detect(Sample.Name, "LiveGas") ~ 101730,
str_starts(Sample.Name, "C-1") ~ 83028,
str_starts(Sample.Name, "C1") ~ 83028,
str_starts(Sample.Name, "TIRI-F") ~ 2138,
str_starts(Sample.Name, "TIRI-I") ~ 17185,
str_starts(Sample.Name, "C-2") ~ 1082,
str_starts(Sample.Name, "C2") ~ 1082,
str_starts(Sample.Name, "NOSAMS") ~ 38809,
str_detect(Sample.Name, "DeadGas") ~ 72446)) %>%
left_join(select(std, rec_num, fm_consensus), by = "rec_num") %>%
mutate(fm_consensus = case_when(rec_num == 101730 ~ 1.0398,
rec_num == 72446 ~ 0.0013,
TRUE ~ fm_consensus))
data %>% distinct(wheel, Pos, Sample.Name, fm_consensus, rec_num) %>% filter(fm_consensus > 0.9)
Current dependent blank with combined data
unique(data$Sample.Name)
Modern
data %>%
filter(!(he12C < 2.5 & normFm > 1.05) & fm_consensus > 0.9) %>%
#filter(fm_consensus > 0.9) %>%
#filter(str_detect(Sample.Name, "LiveGas")) %>%
ggplot(aes(he12C, normFm)) +
#geom_smooth(method = "lm") +
geom_point()
data %>% select(wheel, Pos, Sample.Name, he12C, fm_consensus, normFm) %>% filter(he12C < 2.5 & normFm > 1.05, fm_consensus > 0.9)
Dead
data %>%
#filter(str_detect(Sample.Name, "DeadGas")) %>%
filter(fm_consensus < 0.1 & normFm < 0.3 & Sample.Name != "DilDeadGas6") %>%
ggplot(aes(he12C, normFm)) +
geom_point()
data %>% select(wheel, Pos, Sample.Name, he12C, fm_consensus, normFm) %>% filter(he12C < 1 & normFm < .1, fm_consensus < 0.3)
filter(data, Sample.Name == "DilDeadGas6")
Summarize by sample
ds <- sum_hgis(data) %>%
filter((str_detect(Sample.Name, "LiveGas") & Cur > 1) |
(str_detect(Sample.Name, "DeadGas") & Cur > 1) |
str_detect(Sample.Name, "Blank")) %>%
mutate(Gas = case_when(str_detect(Sample.Name, "LiveGas") ~ "LiveGas",
str_detect(Sample.Name, "DeadGas") ~ "DeadGas",
TRUE ~ "Blank"),
Cur_inv = 1/Cur)
Modern
live <- ds %>%
filter(str_detect(Sample.Name, "LiveGas") & Cur > 1) %>%
ggplot(aes(Cur, mean)) +
# geom_smooth(method = "lm") +
geom_pointrange(aes(ymin = mean - merr, ymax = mean + merr)) +
geom_point()
Dead
dead <- ds %>%
filter(str_detect(Sample.Name, "DeadGas")) %>%
ggplot(aes(Cur, mean)) +
#geom_smooth(method = "lm") +
geom_pointrange(aes(ymin = mean - merr, ymax = mean + merr)) +
geom_point()
filter(ds, Cur < 1)
live / dead
Modelling current dependent blank
Fit linear models
fits <- ds %>%
ungroup() %>%
nest(data = -Gas) %>%
mutate(fit = map(data, ~lm(mean ~ Cur_inv, data = .x)),
tidied = map(fit, tidy)) %>%
unnest(tidied)
blankfit <- fits %>%
select(Gas, term, estimate) %>%
pivot_wider(names_from = c(Gas, term), values_from = estimate) %>%
mutate(inv_m_blank = -(`DeadGas_(Intercept)` - `LiveGas_(Intercept)`)/(DeadGas_Cur_inv - LiveGas_Cur_inv),
Fm_blank = `DeadGas_(Intercept)` + DeadGas_Cur_inv * inv_m_blank,
m_blank = 1/inv_m_blank)
blankfit %>%
select(Fm_blank, m_blank)
blankfit
Fm vs. inverse mass with fits.
ggplot(ds, aes(Cur_inv, mean, color = Gas)) +
geom_abline(slope = blankfit$LiveGas_Cur_inv,
intercept = blankfit$`LiveGas_(Intercept)`) +
geom_abline(slope = blankfit$DeadGas_Cur_inv,
intercept = blankfit$`DeadGas_(Intercept)`) +
geom_pointrange(aes(ymin = mean - merr, ymax = mean + merr), size = .3) +
xlim(0, 10) +
#ylim(0, 1.2)
ggtitle("Fm of live and dead CO2 vs. inverse 12C current",
subtitle = "Fit lines should cross at Fm and current of blank")
Gas standards
Use above data with gas standards normalized to solid OX-I
rownames(ds)
ds %>%
filter(Gas == "LiveGas") %>%
ggplot(aes(x = as.factor(as.numeric(rownames(.))), y = mean)) +
geom_pointrange(aes(ymin = mean - merr, ymax = mean + merr))
2021-05-12 Vial dillution tests
Data
data <- get_hgis_data(here("data/USAMS041521R.txt"), as.Date("2021-05-12"))
Currents and ratios vs time.
data %>%
plot_hgis_time(cor1412he, ce*cor1412he) +
ggtitle("Sample ratio vs. time")
data %>%
plot_hgis_time(he12C) +
ggtitle("Sample current vs. time")
2021-05-14 Carbonate samples
Samples
carb_data <- tribble(~vial, ~name, ~rec_num, ~mass,
1, "C-1_1", 83028, 31.61,
2, "C-1_2", 83028, 34.30,
2, "C-1_3", 83028, 33.17,
4, "TIRI-I_1", 17185, 32.74,
5, "TIRI-I_2", 17185, 31.58,
6, "TIRI-I_3", 17185, 33.25,
7, "NOSAMS-2_1", 38809, 33.31,
8, "NOSAMS-2_2", 38809, 31.89,
9, "NOSAMS-2_3", 38809, 36.47
)
write_csv(carb_data, here("data/carb_data_2021-05-14.csv"))
carb_data
Data
Normalizing to gas standards in pos 25, 26, 36 and 37.
data <- get_hgis_data(here("data/USAMS051421R.txt"), as.Date("2021-05-14")) %>%
mutate(Num = ifelse(Pos %in% c(15, 65), "U", Num))
stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier])
data <- data %>%
mutate(normFm = norm_gas(cor1412he, stdrat))
Currents and ratios vs time.
data %>%
filter(!(Pos %in% 1:4)) %>%
plot_hgis_time(normFm, ce*normFm) +
ggtitle("Sample ratio vs. time")
data %>%
filter(!(Pos %in% 1:4)) %>%
plot_hgis_time(he12C)
ggtitle("Sample current vs time")
Summary of normalized data per sample
std <- getStdTable()
ds <- data %>%
filter(!outlier) %>%
sum_hgis() %>%
mutate(rec_num = case_when(str_starts(Sample.Name, "LiveGas") ~ 101730,
str_starts(Sample.Name, "C1") ~ 83028,
str_starts(Sample.Name, "TIRI-F") ~ 2138,
str_starts(Sample.Name, "TIRI-I") ~ 17185,
str_starts(Sample.Name, "C-2") ~ 1082,
str_starts(Sample.Name, "NOSAMS2") ~ 38809,
str_starts(Sample.Name, "DeadGas") ~ 72446)) %>%
left_join(select(std, rec_num, fm_consensus), by = "rec_num") %>%
mutate(fm_consensus = case_when(rec_num == 101730 ~ 1.0398,
rec_num == 72446 ~ 0.0013,
TRUE ~ fm_consensus))
ds[-(1:4),]
data %>%
filter(as.numeric(Pos) > 4) %>%
ggplot(aes(pos_name, normFm)) +
geom_boxplot(fill = "slategray1") +
theme(axis.text.x = element_text(angle = 45, hjust=1))
Agreement of replicates
Summary statistics to look at reproducibility, accuracy between
targets/vials.
ds %>%
filter(as.numeric(Pos) > 4) %>%
mutate(Name = str_remove(Sample.Name, "_.$")) %>%
group_by(Name, fm_consensus) %>%
summarize(Fm.mean = mean(mean),
Fm.sd = sd(mean),
Cur.mean = mean(Cur),
N = n())
Blank correction
Testing correcting for blank using mass-independent large blank
correction.
Using C-1 with no adjustment for graphite value of blank.
fmblank <- ds %>%
filter(str_starts(Sample.Name, "C1")) %>%
pull(mean) %>%
mean()
fmblank
ds <- ds %>%
mutate(fm_lbc = doLBC(mean, fmblank, 1.0398),
diff = fm_lbc - fm_consensus,
normFm = normFm(fm_lbc, fm_consensus),
sigma = sigma(fm_lbc, fm_consensus, merr))
ds %>%
filter(as.numeric(Pos) > 4) %>%
select(Sample.Name, Cur, fm_consensus, mean, merr, fm_lbc, diff)
ds %>%
write_csv(here("data_analysed/USAMS051421.csv"))
ggplot(ds, aes(fm_consensus, diff)) +
geom_hline(yintercept = 0) +
geom_errorbar(aes(ymin = diff - merr, ymax = diff + merr)) +
geom_point() +
ggtitle("Difference in LBC-corrected vs consensus")
ggplot(ds, aes(fm_consensus, sigma)) +
geom_hline(yintercept = 0) +
geom_point() +
ggtitle("Sigma of LBC-corrected vs consensus")
2021-05-20 Dillution tests
Data
data <- get_hgis_data(here("data/USAMS051421R.txt"), as.Date("2021-05-20"))
Currents and ratios vs time.
data %>%
plot_hgis_time(cor1412he, ce*cor1412he) +
ggtitle("Sample ratio vs. time")
data %>%
plot_hgis_time(he12C) +
ggtitle("Sample current vs. time")
2021-05-28 Pendleton samples
First run of Simon Pendleton's samples, along with C-1, C-2 and modern standard. Reworked data reduction and did analysis in separate file.
2021-06-04 Small(er) sample tests
Tested 5, 10, and 15 mg of C-2 carbonate in vials backfilled with Helium.
Method
- Weigh sample into vial
- Evacuate and acidify as normal with gilson
- Backfill to 1 ATM with He-filled syringe 2h after acidification.
Allowed vacuum in vial to pull in He from syringe, then added 1 additional mL He. Should calculate volume CO2 in vial and measure He added to check that things make sense.
Used same capillaries as usual, but decreased He displacement flow to 140 ul/min. Source capillary flow assumed to be around 30ul/min.
Samples
carb_data <- tribble(~number, ~name, ~rec_num, ~mass,
1, "C-2_1", 1082, 6.03,
2, "C-2_2", 1082, 11.10,
3, "C-2_3", 1082, 15.52
)
write_csv(carb_data, here("data/carb_data_2021-06-04.csv"))
carb_data
Data
df <- process_hgis_results(here("data/USAMS052621R.txt"), as.Date("2021-06-04")) #, standards)
Currents and ratios vs time.
df %>%
filter(!(pos %in% 1:4)) %>%
plot_hgis_time(norm_ratio, sig_norm_ratio) +
ggtitle("Sample ratio vs. time")
df %>%
filter(!(pos %in% 1:4)) %>%
plot_hgis_time(he12C) +
ggtitle("Sample current vs time")
Results
Note: no blanks run, so need to use average blank if LBC needed.
df_sum <- sum_hgis_targets(df) %>%
norm_hgis()
df_sum %>%
filter(as.numeric(pos) > 4) %>%
select(pos, sample_name, he12C, n_runs, norm_ratio, sig_norm_ratio) %>%
arrange(sample_name)
df_sum %>%
filter(as.numeric(pos) > 4) %>%
ggplot(aes(sample_name, norm_ratio)) +
geom_pointrange(aes(ymin = norm_ratio - sig_norm_ratio, ymax = norm_ratio + sig_norm_ratio),
size = .5) +
theme(axis.text.x = element_text(angle = 45, hjust=1)) +
labs(x = NULL,
y = "Fraction Modern")
Agreement of replicates
Compare results for samples with more than one replicate.
df_sum %>%
filter(as.numeric(pos) > 4) %>%
compare_replicates()
Agreement with consensus
df_sum %>%
filter(as.numeric(pos) > 4,
!is.na(fm_consensus)) %>%
select(sample_name, fm_consensus, norm_ratio, sig_norm_ratio) %>%
mutate(Fm_diff = norm_ratio - fm_consensus,
sigma = amstools::sigma(norm_ratio, fm_consensus, sig_norm_ratio)) %>%
arrange(sample_name)
2021-06-11 Small(er) sample tests
Tested 8mg of C-1, C-2, and NOSAMS2 carbonate in vials backfilled with Helium.
Method
- Weigh sample into vial
- Evacuate and acidify as normal with gilson
- Backfill to 1 ATM with He-filled syringe 2h after acidification.
Allowed vacuum in vial to pull in He from syringe, then added 1 additional mL He. Should calculate volume CO2 in vial and measure He added to check that things make sense.
Used same capillaries as usual, but decreased He displacement flow to 140 ul/min. Source capillary flow assumed to be around 30ul/min.
Samples
carb_data <- tribble(~name, ~rec_num, ~carbonate_mass, ~helium_added,
"C-1_1", 83028, 8.29, 5,
"C-1_2", 83028, 9.01, 5,
"C-1_3", 83028, 9.00, 5,
"C-2_1", 1082, 8.23, 5,
"C-2_2", 1082, 8.77, 5,
"C-2_3", 1082, 7.34, 5,
"NOSAMS-2_1", 38809, 7.72, 5,
"NOSAMS-2_2", 38809, 7.69, 5,
"NOSAMS-2_3", 38809, 8.86, 5
)
write_csv(carb_data, here("data/carb_data_2021-06-04.csv"))
#insert_samples(carb_data)
carb_data
Data
df <- process_hgis_results(here("data/USAMS061121R.txt"), as.Date("2021-06-11")) #, standards)
#insert_raw(df)
Currents and ratios vs time.
df %>%
filter(pos > 2) %>%
plot_hgis_time(norm_ratio, sig_norm_ratio, outlier = !.$ok_calc) +
ggtitle("Sample ratio vs. time")
df %>%
filter(pos > 2) %>%
plot_hgis_time(he12C, outlier = !.$ok_calc) +
ggtitle("Sample current vs time")
Results
Note: no blanks run, so need to use average blank if LBC needed.
df_sum <- sum_hgis_targets(df) %>%
norm_hgis() %>%
blank_cor_hgis()
#insert_results(df_sum)
df_sum %>%
filter(pos > 2) %>%
mutate(he12C = he12C*1E6) %>%
select(Positon = pos, `Sample Name` = sample_name, `12C Current (uA)` = he12C, n_runs, fm_corr, sig_fm_corr) %>%
arrange(`Sample Name`)
df_sum %>%
plot_hgis_summary()
Agreement of replicates
Compare results for samples with more than one replicate.
df_sum %>%
filter(pos > 2) %>%
compare_replicates()
Agreement with consensus
cons <- df_sum %>%
filter(pos > 2,
!is.na(fm_consensus)) %>%
select(sample_name, fm_consensus, fm_corr, sig_fm_corr) %>%
mutate(fm_diff = fm_corr - fm_consensus,
sigma = amstools::sigma(fm_corr, fm_consensus, sig_fm_corr)) %>%
arrange(sample_name)
cons
knitr::kable(cons)
blongworth/HybridGIS documentation built on Dec. 19, 2021, 10:41 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
# Load libraries and functions library(tidyverse) library(amstools) library(here) library(HybridGIS) library(broom) library(patchwork) theme_set(theme_bw()) options(digits = 3)
Development notes for using the HGIS with an open split. Intention is to use with gasbench for carbonates and/or large co2 samples.
Capillary parameters
Calculate ideal capillary to provide \~7.5uL/min at atmospheric pressure.
Short capillary at 1 atm
40um capillary
r <- 2E-5 u <- 1.49E-5 x <- 3.5 flowcalc(100, r, u, x)
r <- 2E-5 u <- 1.49E-5 x <- 2 flowcalc(100, r, u, x)
50um capillary
r <- 2.5E-5 u <- 1.49E-5 x <- 3 flowcalc(100, r, u, x)
r <- 2.5E-5 u <- 1.49E-5 x <- 7 flowcalc(100, r, u, x)
First Tests
2018-04-10 - First tests
Tried a bunch of capillary combinations. Other than initial pressure burst when opening or closing HGIS valve, couldn't see any source pressure changes or beam current. We believe the source end of the stainless capillary was clogged by cesium residue.
Cleaned capillary end at next source cleaning, first with kimwipe and water, then with isopropanol. Lots of black crud. Confirmed open passage by blowing N2 through capillary into cup of isopropanol to observe bubbles.
2018-05-23 - First tests
Tried 3m of 50um glass capillary. Insulated from source to cage with teflon tubing as outer jacket. No problems with sparks.
Low currents, even when cutting capillary down to 2m. Either steel capillary is partially clogged or tip is not making good seal with target.
2018-05-31 - He Dilution
Open split set up with helium dilution. Using 100um glass capillary. \~50cm source to bulkhead with teflon jacket, 285cm bulkhead to open split.
Flows
flowcalc(100, r = 2.5E-5, u = 1.49E-5, l = 3.35)
This is way too high. I may have been using a 50um capillary for the tests.
Pressures
| Condition | Torr | |--------------|-------| | base w/cryo | 1E-7 | | base w/turbo | 4E-7 | | capillary | 36E-7 |
: Source pressures
Dilution and current
Flow and current tests with varying capillary and dilution. Flows in mL/min, pressure in Torr, current in uA. Replumbed helium to panel open split with needle valve partway through to better control flow. Live CO2 set to 10mL/min.
osdil <- read_csv(here("data/2018-05-31_dillution.csv")) %>% mutate(ratio = helium/co2) osdil
osdil %>% filter(caplen == 322) %>% ggplot(aes(ratio, current)) + geom_line() + ggtitle("Current vs dillution ratio for 10 and 18 mL/m CO2") + facet_grid(co2 ~ .)
Test Acquisitions
Ran 4 test Acquisitions with open split: pure CO2, pure dead CO2, diluted live, diluted dead. Data are 14/12 ratio with a 13C correction applied. NormFm is values normalized to the ratio of the solid OX-I. There are also dual inlet samples on this wheel, filtered out by date. Positions are incorrect in file because wheel was advanced manually. Correct positions in notebook and fixed below. Added correct dilution factors below.
data <- get_hgis_data(here("data/USAMS053018R.txt"), as.Date(c("2018-05-30", "2018-05-31"))) %>% filter(Pos %in% c(3,4) | as.Date(ts) == "2018-05-31") %>% mutate(Pos = case_when(Sample.Name == "LiveGas" ~ 7, Sample.Name == "DeadGas" ~ 8, Sample.Name == "LiveGasHe" ~ 9, Sample.Name == "DeadGasHe" ~ 10, TRUE ~ as.numeric(Pos)), dil_factor = case_when(Pos == 9 ~ (50.3 - 9.4)/9.4, Pos == 10 ~ (40 - 4.8)/4.8), Num = ifelse(Pos %in% c(3, 4), "S", Num)) # Get mean ratio for solid OX-I and normalize data stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data <- mutate(data, normFm = norm_gas(cor1412he, stdrat)) sum_hgis(data)
plot_hgis(data)
2019-05-31 - startup tests
Reinstalled gas delivery arm after cleaning and polishing ball end.
Using 50um x .5m capillary source to cage and 50um x 2.5m to open split. Open split flowing modern co2 at 10mL/min. Using same targets and wheel as last time. Surprisingly good performance, including no-gas base current, hgis current and solid samples.
flowcalc(pres = 100, r = 2.5E-5, u = 1.49E-5, l = 3.0)
Currents up to 15uA, source vac 8.3E-7 with gas on, 5.5E-7 with gas off.
No data recorded
2019-06-03 - dillution tests
Spun up inlet turbopump. Looked at dilution and current.
dil <- tibble( CO2 = c(10, 10, 10, 10, 10), Tot = c(10, 20, 30, 43.5, 10), vac = c(1.24E-6, 9.52E-7, 8.3E-7, 8E-7, 1.07E-6), cur = c(10, 9.6, 7, 5, 10) ) %>% mutate(He = Tot - CO2, ratio = He/CO2) dil
ggplot(dil, aes(ratio, cur)) + geom_line() + ggtitle("Current vs dillution ratio, CO2 at 10ml/min")
Parameterization of source and open split
2020-09-10 Starting up HGIS
Solid run on pos 1:4 on USAMS012220.
2020-09-29 Dilution vs Ratio
Testing effect of helium dilution on ratio. Used remaining targets on USAMS012220. Reterminated capillaries at cage wall and at open split.
Used positions 7-10, results in USAMS012220. Low currents: \~6uA pure CO2, 3uA diluted. Flows: He 10, deadCO2 11, liveCO2 10.
# Load data data <- get_hgis_data(here("data/USAMS012220R.txt"), as.Date(c("2020-09-10", "2020-09-29"))) %>% mutate(dil_factor = ifelse(str_starts(Sample.Name, "Dil"), 1, 0), Num = ifelse(Pos == 7, "U", Num)) stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data <- mutate(data, normFm = norm_gas(cor1412he, stdrat)) sum_hgis(data)
plot_hgis(data)
Looks like diluted and undiluted live gas have the same ratio.
2020-10-15 New wheel, low currents
Loaded a new wheel (USAMS101320) with 4 solid targets and 6 gas targets. Currents were low (3uA gas, 40uA solid) and didn't increase with tune, target position or gas arm position. Will clean source and try again.
2020-11-03 Continued dillution tests
Still very low currents, \~2uA for live CO2. Ran a dillution test anyway with positions 5-8 on USAMS101320.
Flows
| Gas | Flow | |------|------| | live | 10 | | dead | 11.1 | | He | 30.2 |
data <- get_hgis_data(here("data/USAMS101320R.txt"), as.Date("2020-11-03")) sum_hgis(data)
plot_hgis(data)
2020-11-17 Dillution testing - back to good currents
Finally back to testing after figuring out current issue. Rerunning dilution test. Flows: live 11.6mL/min, dead 11.1, He 30.2. Pressures: cryo base 8E-8, turbo base 3E-7, w/live CO2 1.2E-6. Source producing \~8uA HE12C with live CO2. Data in USAMS101320.
Running positions 9 and 10 as 1:1 He dilutions of live and dead, respectively.
data <- get_hgis_data(here("data/USAMS101320R.txt"), as.Date("2020-11-17")) sum_hgis(data)
Look at counts per second to think about needed acquisition time.
data %>% group_by(Sample.Name) %>% summarize(cps = mean(CntTotGT/(Cycles/10)))
plot_hgis(data)
data %>% filter(Pos %in% 5:10) %>% ggplot(aes(he12C, normFm, color = Sample.Name)) + geom_point()
Doesn't seem to follow curves expected for a MBC with nearly live blank at about 1% of gas current.
2020-12-04 Blank tests
Ran a series of blank tests before exposing graphite samples or introducing live co2.
Capillary break between Target blank and Helium blank. Replaced with SGE 100um x 52cm. Initial currents of targets are high, between 10-25uA. Drops asymptotically to base target current in 5-10 minutes.
data <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2020-12-04")) sum_hgis(data)
plot_hgis(data)
Modeling Blank
With these data we can start to think about modeling a mass balance, or more accurately, a current balance blank. It's not initially clear whether the blank comes from the target/source, diluent helium, or a combination of the two.
data %>% filter(Pos %in% 9:12) %>% ggplot(aes(he12C, normFm, color = Sample.Name)) + geom_point() + ggtitle("Blank Fm vs. current")
If this blank follows the constant contamination model, plotting Fm vs. inverse current should give a linear fit. Need more data here to confirm fit and determine current and Fm of blank.
# subset data gas <- data %>% filter(Pos %in% 9:12) %>% mutate(icur = 1/he12C) # fit a linear model fitlm <- lm(normFm ~ icur, data = gas) # Add predicted values to data gas$Fm = predict(fitlm) ggplot(gas) + geom_line(aes(1/he12C, y = Fm)) + geom_point(aes(1/he12C, normFm, color = Sample.Name)) + ggtitle("Blank Fm vs. inverse current")
TODO: Fit mass balance curve using currents TODO: Use bottle gas to produce live/dead curves TODO: produce live/dead curves using helium dilution
2020-12-11 open split flow reduction
Testing reduced open split flows. Worked with flow controller (helium) and adjusted stops on live and dead co2 needle valves to allow lower flows. Tested for breakthrough by flowing helium through split and into source while blowing dead CO2 across the opening of the split at \~10mL/min. CO2 seen in source somewhere between 100 and 300 uL/min. Hard to measure with this setup due to slow response time and lack of sensitivity of capillary/ion source. Need to try this with leak checker.
Reran blank tests at lower flows. Results seem as good or better than last week's tests at high flows. This may also be due to targets being more pumped down or burned in.
TODO: get flows from notes.
data <- get_hgis_data(here("data/USAMS120320R.txt")) %>% filter(Pos %in% 1:4 | as.Date(ts) == "2020-12-11") %>% mutate(he = case_when(Pos == 9 ~ (1059 - 383), Pos == 10 ~ 0, Pos == 11 ~ 1370, Pos == 12 ~ 0), co2 = case_when(Pos == 9 ~ 383, Pos == 10 ~ 383, Pos == 11 ~ 0, Pos == 12 ~ 0), dil_rat = he/co2) data %>% group_by(Pos, Sample.Name, dil_rat) %>% summarise(Cur = mean(he12C), Cur.sd = sd(he12C), mean = mean(normFm), sd = sd(normFm), interr = 1/sqrt(sum(CntTotGT)), acqtime = sum(Cycles)/10, N_acq = n())
plot_hgis(data)
# subset data gas <- data %>% filter(Pos %in% 9:12) %>% mutate(icur = 1/he12C) # fit a linear model fitlm <- lm(normFm ~ icur, data = gas) # Add predicted values to data gas$Fm = predict(fitlm) ggplot(gas) + geom_line(aes(1/he12C, Fm)) + geom_point(aes(1/he12C, normFm, color = Sample.Name)) + ggtitle("Blank Fm vs. inverse current")
2020-12-18 More testing of low open split flow rates
Tested open split breakthrough at low flows using leak checker. Measured helium flow through capillary using leak checker. Ran target blanks, dead and live samples at low flows. Started building double needle testbed.
Leak checker flow rate
Connected cage-split capillary to leak checker. Base leak rate with capillary in air: 1.7E-9 mbarLs-1. Leak rate with helium in open split: 3.5E-4 mbarls-1.
Leak checker breakthrough tests
Flowing dead CO2 through split. Flowing helium across mouth of split at 8mLmin-1. Detection of helium means atmosphere is getting into split.
lc_breakthrough <- tribble(~co2, ~leakrate, 730, 6E-9, 735, 6.5E-9, 525, 7.9E-9, 178, 1.3E-7, 150, 2E-7, 0, 3E-6) lc_breakthrough ggplot(lc_breakthrough, aes(co2, leakrate)) + geom_line() + geom_point() + ggtitle("Helium leak rate vs. CO2 flow to split")
Data acquisitions at low(er) flows
Source off base: 4E-8 Cryo base (HV + Ion): 6E-8 Turbo base: 2E-7
Set SRS to 10uA/V and 50nA/V for 12 and 13C.
Flows:
flows <- tribble(~pos, ~Sample.Name, ~co2, ~he, 10, "DeadGas", 270, 0, 9, "DilDeadGas", 112, 288, 8, "DilDeadGas6", 102, 0, 8, "DeadGas", 102, 572-102, 7, "HeliumBlank", 0, 572-102, 7, "DilLiveGas6", 117, 576-117) %>% mutate(dil_ratio = he/co2) flows
Data
data <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2020-12-18")) sum_hgis(data)
plot_hgis(data)
data %>% filter(Pos %in% 7:10) %>% ggplot(aes(1/he12C, normFm, color = Sample.Name)) + geom_point() + ggtitle("Fm of live and dead samples vs. inverse current")
2021-01-08 Pure gas precison testing
Plumbed long 50um capillary from live CO2 regulator to cage wall fitting. This allows controlling flow of CO2 to source using regulator pressure.
Current vs flow and pressure
First test was with 748cm x 50 um capillary. Source base pressure using the cryopump was 2E-7 Torr. Base current for Pos 7 was 300uA. Initial max current \@ 200kPA CO2 was 9uA. Did not return to this level in the pressure test following, not sure why, source pressure shows CO2 was getting to source. Typically best currents with pure CO2 are at a source pressure of around 1E-6 Torr. Reg pressure in kPa, source in Torr x 10-7 and current in uA.
cur_vs_pres <- tribble(~regp, ~sourcep, ~c12cur, 0, 2.8, 0.8, 50, 3.3, 1, 100, 5.3, 1.2, 150, 8.2, 1.4, 200, 12, 1.8, 250, 17, 2.4, 300, 22, 3.1, 350, 28, 3.7, 400, 35, 4.5) cur_vs_pres
Tested again with new capillaries after a clog in regulator-cage capillary and spark damage to cage-source capillary. New capillaries: reg-cage- 639cm x 50um, cage-source- 50cm x 50um. Base current 250nA.
cur_vs_pres2 <- tribble(~regp, ~sourcep, ~c12cur, 25, 2, .25, 100, 4, 1.3, 150, 8.4, 6.3, 200, 12, 6.2, 250, 17, 6.2) cur_vs_pres2
Data vs flow
Took data on targets 6 and 7 at various CO2 flow rates.
Data summary
data <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2021-01-08")) ds <- sum_hgis(data) ds
plot_hgis(data)
Ratio and current vs flow
Took data for 4 different flows on target 6. CO2 flow data show ratio dependence on flow. Need to fill in data to get shape of curve. Current vs flow
ds %>% filter(Pos == 6) %>% mutate(Pressure = as.numeric(str_extract(Sample.Name, "\\d+"))) %>% ggplot(aes(Pressure, Cur)) + geom_pointrange(aes(ymin = Cur - Cur.sd, ymax = Cur + Cur.sd)) + geom_point(size = 4) + ggtitle("Fm of live CO2 vs. C12 current")
Fm vs. Flow. Note that solid standards had a lot of variability, so treat ratio as relative, not well normalized.
ds %>% filter(Pos == 6) %>% mutate(Pressure = str_extract(Sample.Name, "\\d+")) %>% ggplot(aes(Cur, mean, shape = Pressure, color = Pressure)) + geom_pointrange(aes(ymin = mean - sd, ymax = mean + sd)) + geom_point(size = 4) + ggtitle("Fm vs. C12 current for varying CO2 regulator pressure")
2021-01-15 Precision and vial tests
Used test vial setup to test pure CO2 precision and introduction of CO2 from vial by displacing with helium.
Data
df <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2021-01-15")) stdrat <- mean(df$cor1412he[df$pos_name == "8 - LiveGasOS" & !df$outlier]) df <- mutate(df, normFm = norm_gas(cor1412he, stdrat)) sum_hgis(df)
Ratio and current during runs
plot_hgis_time(df) + facet_wrap(vars(pos_name), scales = "fixed") ggplot(df, aes(cum_acqtime, he12C )) + geom_point() + facet_wrap(vars(pos_name), scales = "fixed")
Ratio dependence on current
ggplot(df, aes(he12C, normFm, color = pos_name)) + geom_point() + ggtitle("Fm of modern CO2 vs. C12 current")
ggplot(df, aes(1/he12C, normFm, color = pos_name)) + geom_point() + ggtitle("Fm of modern CO2 vs. inverse C12 current")
Question: Is ratio current depletion change in CO2 from vial or target degradation? Long pure CO2 run to test.
2021-01-22 Pure gas precison testing
Plumbed long 50um capillary from Valco valve to cage wall fitting. Plumbed dead CO2 to Valco with 1/16" ss.
Took data with dead CO2 on target 5 at various CO2 flow rates. Same with liveCO2 on target 6. Source not performing well, low currents and current did not correspond well to CO2 pressure/flow. Current vs ratio data may still be useful for blank model. Solid currents good, so this is a gas issue. Gas blank also not great.
Data summary
data <- get_hgis_data(here("data/USAMS012121R.txt"), as.Date("2021-01-22")) ds <- sum_hgis(data) ds
Note that solid standards had a lot of variability, so treat ratio as relative, not well normalized.
plot_hgis(data)
Relationship of current and ratio to CO2 flow rate
Current vs flow
Supply pressure is proportional to flow rate.
ds <- ds %>% filter(Pos %in% c(5, 6)) %>% mutate(Cur_inv = 1/Cur, Pressure = as.numeric(str_extract(Sample.Name, "\\d+")), Gas = ifelse(Pos == 5, "DeadCO2", "LiveCO2")) ggplot(ds, aes(Pressure, Cur, color = Gas)) + geom_pointrange(aes(ymin = Cur - Cur.sd, ymax = Cur + Cur.sd)) + geom_point(size = 4) + ggtitle("C12 current vs. CO2 regulator pressure")
Fm vs. current and flow rate
ggplot(ds, aes(Cur, mean, color = Pos)) + geom_pointrange(aes(shape = factor(Pressure), ymin = mean - sd, ymax = mean + sd)) + ggtitle("Fm of live and dead gas vs C12 current")
Modelling current dependent blank
Fit linear models
fits <- ds %>% ungroup() %>% nest(data = -Gas) %>% mutate(fit = map(data, ~lm(mean ~ Cur_inv, data = .x)), tidied = map(fit, tidy)) %>% unnest(tidied) fits %>% select(Gas, term, estimate) %>% pivot_wider(names_from = c(Gas, term), values_from = estimate) %>% mutate(inv_m_blank = -(`DeadCO2_(Intercept)` - `LiveCO2_(Intercept)`)/(DeadCO2_Cur_inv - LiveCO2_Cur_inv), Fm_blank = `DeadCO2_(Intercept)` + DeadCO2_Cur_inv * inv_m_blank, m_blank = 1/inv_m_blank) %>% select(Fm_blank, m_blank)
Fm vs. inverse mass with fits.
ggplot(ds, aes(Cur_inv, mean, color = Pos)) + geom_pointrange(aes(shape = factor(Pressure), ymin = mean - sd, ymax = mean + sd)) + geom_smooth(method = "lm", fullrange = TRUE) + xlim(0, 5) + ggtitle("Fm of live and dead CO2 vs. inverse 12C current", subtitle = "Fit lines should cross at Fm and current of blank")
For these measurements, it looks like the blank is roughly 225nA at half modern. This is higher current and lower Fm than expected based on previous measurements. This is in line with the high blanks seen when measuring dead carbon.
Gas delivery from vials
2021-01-29 Reproducibility between multiple vials
Measured live and dead CO2 from vials. "STD" samples were displaced to split with flowing CO2, while the "V" samples were displaced by helium as they would be for unknowns. All samples measured on fresh targets after a 5 minute "burn in" to reduce blank.
Gas flows
Replaced capillary from Valco valve to double needle with longer piece of 50um capillary. Set helium and CO2 flows.
+----------------------+-----------------------+-----------------------+ | Gas | Pressure (kPa) | Flow (uL/min) | +======================+=======================+=======================+ | LiveCO2 | 165 | 136 | +----------------------+-----------------------+-----------------------+ | DeadCO2 | 160 | 120 | +----------------------+-----------------------+-----------------------+ | Helium | 165 | 115 | +----------------------+-----------------------+-----------------------+
: Gas flow rates
Measured split to cage wall capillary flow with leak checker. Leak rate: 1.0E-5 Pa m\^3/s.
Data
Normalizing to solid OX-I's.
data <- get_hgis_data(here("data/USAMS012121R.txt"), as.Date("2021-01-29")) %>% mutate(Num = ifelse(Pos %in% c(7, 9), "U", Num))
Currents and ratios vs time.
Sample ratio vs. time
plot_hgis_time(data, cor1412he, ce*cor1412he)
Sample current vs time
plot_hgis_time(data, he12C)
# Get mean ratio for solid OX-I and normalize data stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data <- mutate(data, normFm = norm_gas(cor1412he, stdrat))
Per-sample summary
ds <- data %>% filter(!outlier) %>% sum_hgis() ds
plot_hgis(data)
Agreement of replicates
Summary statistics to look at reproducibility between targets/vials.
ds %>% filter(as.numeric(Pos) > 4, Pos != 9) %>% # Don't use bad target mutate(gas = ifelse(str_starts(Sample.Name, "Live"), "Live", "Dead"), helium = ifelse(str_detect(Sample.Name, "V\\d$"), TRUE, FALSE)) %>% group_by(gas, helium) %>% summarize(Fm.mean = mean(mean), Fm.sd = sd(mean), N = n())
It looks like the SD of modern samples is around 1%, and that pure CO2 performs slightly better than helium displacement.
2021-02-05 Carbonate samples
Measured carbonate samples prepared on gasbench. Samples included C-1, TIRI-F, TIRI-I, C-2, and NOSAMS2. Normalized to gas standards or solid OX-I.
First carbonate sample clogged capillary. Capillary replaced, moved needle up to avoid sucking in acid. Was not able to get good currents after replacing capillary, even trying several length/capillary combinations. Source vacuum indicates normal amount of gas entering source. May need to consider a water trap. All samples measured on fresh targets after a 5 minute "burn in" to reduce blank.
Samples
carb_data <- tribble(~vial, ~name, ~rec_num, ~mass, 1, "C-1", 83028, 32.05, 2, "TIRI-F", 2138, 33.56, 3, "TIRI-I", 17185, 32.67, 4, "TIRI-I", 17185, 32.35, 5, "C-2", 1082, 31.35, 6, "C-2", 1082, 34.11, 7, "NOSAMS2", 38809, 32.30, 8, "NOSAMS2", 38809, 34.00 ) write_csv(carb_data, here("data/carb_data_2021-02-05.csv")) carb_data
Gas flows
Set helium and CO2 flows.
+----------------------+-----------------------+-----------------------+ | Gas | Pressure (kPa) | Flow (uL/min) | +======================+=======================+=======================+ | LiveCO2 | 150 | 95 | +----------------------+-----------------------+-----------------------+ | Helium | 150 | 97 | +----------------------+-----------------------+-----------------------+
: Gas flow rates
Data
Normalizing to gas standards in pos 5 and 8.
data <- get_hgis_data(here("data/USAMS020521R.txt")) %>% mutate(Num = ifelse(Pos %in% c(2, 4), "U", Num)) stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat))
Currents and ratios vs time.
Sample ratio vs. time
plot_hgis_time(data, normFm, ce*normFm)
Sample current vs time
plot_hgis_time(data, he12C)
Summary of normalized data per sample
ds <- data %>% filter(!outlier) %>% sum_hgis() %>% mutate(rec_num = case_when(str_starts(Sample.Name, "LiveGas") ~ 101730, Sample.Name == "C-1" ~ 83028, Sample.Name == "TIRI-F" ~ 2138, Sample.Name == "TIRI-I" ~ 17185, Sample.Name == "C-2" ~ 1082, Sample.Name == "NOSAMS2" ~ 38809)) std <- getStdTable() ds <- left_join(ds, select(std, rec_num, fm_consensus)) %>% mutate(fm_consensus = ifelse(rec_num == 101730, 1.0398, fm_consensus)) ds[-(1:4),-3]
data %>% filter(as.numeric(Pos) > 4) %>% ggplot(aes(pos_name, normFm)) + geom_boxplot(fill = "slategray1") + theme(axis.text.x = element_text(angle = 45, hjust=1))
Agreement of replicates
Summary statistics to look at reproducibility, accuracy between targets/vials.
ds %>% filter(as.numeric(Pos) > 4, Pos != 6) %>% # Don't use bad target mutate(Name = ifelse(Pos %in% c(5,8), "LiveGas", Sample.Name)) %>% group_by(Name, fm_consensus) %>% summarize(Fm.mean = mean(mean), Fm.sd = sd(mean), Cur.mean = mean(Cur), N = n())
It looks like the SD of replicates of modern samples with current > 3uA is around 0.5%, which is better than measurement statistics. Fm of NOSAMS 2 is a bit high. Need to investigate current-dependent blank.
Blank correction
Testing correcting for blank using a simple, mass-independent large blank correction.
LBC does not work well for this set of samples.
fmblank <- ds$mean[ds$Sample.Name == "TIRI-F"] fmblank ds <- ds %>% mutate(fm_lbc = doLBC(mean, fmblank, 1.0398)) ds %>% filter(as.numeric(Pos) > 4) %>% select(Sample.Name, Cur, fm_consensus, mean, fm_lbc) ds %>% write_csv(here("data_analysed/USAMS020521.csv"))
2021-03-01 Live gas and solid samples
Ran live CO2 from vial to check performance after source cleaning and cryopump issues. Slow to get going, so only one modern gas standard and solid samples run. Tried second modern gas standard on pos 16, but no current from gas that worked on pos 15. Bad target?
data <- get_hgis_data(here("data/USAMS020521R.txt"), as.Date("2021-03-01"))
Currents and ratios vs time.
Sample ratio vs. time
plot_hgis_time(data, normFm, ce*normFm)
Sample current vs time
plot_hgis_time(data, he12C)
Normalization
# Get mean ratio for solid OX-I and normalize data stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat))
Summary of normalized data per sample
ds <- data %>% filter(!outlier) %>% sum_hgis() %>% mutate(rec_num = case_when(str_starts(Sample.Name, "LiveGas") ~ 101730, Sample.Name == "C-1" ~ 83028, Sample.Name == "TIRI-F" ~ 2138, Sample.Name == "TIRI-I" ~ 17185, Sample.Name == "C-2" ~ 1082, Sample.Name == "NOSAMS2" ~ 38809)) std <- getStdTable() ds <- left_join(ds, select(std, rec_num, fm_consensus)) %>% mutate(fm_consensus = ifelse(rec_num == 101730, 1.0398, fm_consensus)) ds
plot_hgis(data)
2021-03-05 Carbonate samples
Measured carbonate samples prepared on gasbench. Samples included C-1, C-2, and NOSAMS2. Normalized to gas standards or solid OX-I.
All samples measured on fresh targets after a 5 minute "burn in" to reduce blank. Gas flows same as 2021-02-05.
Samples
carb_data <- tribble(~vial, ~name, ~rec_num, ~mass, 1, "C-1", 83028, 33.69, 2, "C-1", 83028, 31.62, 3, "C-2", 1082, 32.70, 4, "C-2", 1082, 32.08, 5, "NOSAMS2", 38809, 33.23, 6, "NOSAMS2", 38809, 33.95 ) write_csv(carb_data, here("data/carb_data_2021-03-05.csv")) carb_data
Data
Normalizing to gas standards in pos 5 and 8.
data <- get_hgis_data(here("data/USAMS030421R.txt"), as.Date("2021-03-05"))
Currents and ratios vs time.
Sample ratio vs. time
plot_hgis_time(data, normFm, ce*normFm)
Sample current vs time
plot_hgis_time(data, he12C)
Summary of normalized data per sample
std <- getStdTable() ds <- data %>% filter(!outlier) %>% sum_hgis() %>% mutate(rec_num = case_when(str_starts(Sample.Name, "LiveGas") ~ 101730, str_starts(Sample.Name, "C1") ~ 83028, str_starts(Sample.Name, "TIRI-F") ~ 2138, str_starts(Sample.Name, "TIRI-I") ~ 17185, str_starts(Sample.Name, "C2") ~ 1082, str_starts(Sample.Name, "NOSAMS2") ~ 38809, str_starts(Sample.Name, "DeadGas") ~ 72446)) %>% left_join(select(std, rec_num, fm_consensus), by = "rec_num") %>% mutate(fm_consensus = case_when(rec_num == 101730 ~ 1.0398, rec_num == 72446 ~ 0.0013, TRUE ~ fm_consensus)) ds[-(1:4),-3]
data %>% filter(as.numeric(Pos) > 4) %>% ggplot(aes(pos_name, normFm)) + geom_boxplot(fill = "slategray1") + theme(axis.text.x = element_text(angle = 45, hjust=1))
Agreement of replicates
Summary statistics to look at reproducibility, accuracy between targets/vials.
Mean of modern CO2 filled on HGIS setup (LiveGas) and Gilson (LiveGasGil) were used for normalization but were fairly different on this run, so shown separately.
ds %>% filter(as.numeric(Pos) > 4) %>% mutate(Name = str_remove(Sample.Name, "-.$")) %>% group_by(Name, fm_consensus) %>% summarize(Fm.mean = mean(mean), Fm.sd = sd(mean), Cur.mean = mean(Cur), N = n())
Agreement of replicates is better than 2% for all samples. Samples have higher Fm than consensus. I think most of this is due to blanks, but the low gas standard also would push unknown Fm's up.
Blank correction
Testing correcting for blank using a simple, mass-independent large blank correction.
Using dead gas blank, naive with no adjustment for graphite value of blank. This pushes Fm in the right direction, but comparing samples of the same type, there's a lot of variability that seems to be tied to sample current.
fmblank <- ds$mean[ds$Sample.Name == "DeadGas"] fmblank ds <- ds %>% mutate(fm_lbc = doLBC(mean, fmblank, 1.0398)) ds %>% filter(as.numeric(Pos) > 4) %>% select(Sample.Name, Cur, fm_consensus, mean, fm_lbc) ds %>% write_csv(here("data_analysed/USAMS030421.csv"))
2021-03-26 High dillution test
Determine length of 75um capillary for \~100ulm to source.
r <- 7.5E-5 / 2 u <- 1.49E-5 x <- 3 flowcalc(100, r, u, x)
Startup and test after opening tank to refill terminal argon bottle. Testing helium dilution by starting with helium in vial and displacing with CO2. Used 2.5m of 75um capillary to provide 100ulm to source, but initial source pressure of 3E-7 Torr indicates lower flows. Shortened split-cage capillary to about 150cm, without a huge effect on pressure. Flow may be defined by cage-source capillary.
LiveCO2 flow set to 244 uL/m.
Using position 14 on USAMS030421. Initial current after 10min: 150nA.
Currents rose to a stable 10uA and stayed there for more than 90 min!
Data
data <- get_hgis_data(here("data/USAMS030421R.txt"), as.Date("2021-03-26")) %>% mutate(Num = ifelse(Pos %in% c(2,4), "S", Pos)) stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat))
Currents and ratios vs time.
data %>% filter(Pos == 14) %>% plot_hgis_time(normFm, ce*normFm) + ggtitle("Sample ratio vs. time")
data %>% filter(Pos == 14) %>% plot_hgis_time(he12C) + ggtitle("Sample current vs. time")
Predicted flow of CO2 to source for test flow conditions
co2flow <- 244 / 60 #flow in ul/s data.frame(x = 0:8000) %>% ggplot(aes(x)) + stat_function(fun=function(x) concCO2(x, r = co2flow) * 7, aes(color = "He")) + stat_function(fun=function(x) 7 - concCO2(x, r = co2flow) * 7, aes(color = "CO2")) + scale_color_manual("Gas", values = c("green", "blue")) + labs(title = "CO2 in vial over time", subtitle = "7mL vial, 244uL/min CO2", x = "Time (s)", y = "Volume (mL)")
Stable currents reached at around 5mL CO2 in vial, but stable ratios reached at around 2mL. Contrast with helium dilution test from 4/2 below.
Ratio vs Current
data %>% filter(Pos == 14) %>% ggplot(aes(he12C, normFm)) + geom_point()
2021-04-02 High dillution test
Testing performance at higher dilution ratio. Started with pure live CO2 in vial and did long acquisition with high displacement helium flow to check current as CO2 concentration decreases.
Helium flow set to 244 uL/m.
Using position 25 on USAMS040321. Shifted targets to higher positions to avoid wearing only low wheel positions. Initial current after 10min: 150nA.
Data
data <- get_hgis_data(here("data/USAMS040121R.txt"), as.Date("2021-04-02")) %>% mutate(Num = ifelse(Pos == 5, "S", Pos)) stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat))
Currents and ratios vs time.
Sample ratio vs. time
data %>% filter(Pos == 5) %>% plot_hgis_time(normFm, ce*normFm)
Sample current vs time
data %>% filter(Pos == 5) %>% plot_hgis_time(he12C)
Expected CO2 ratio/flow
Plot volume of CO2 in sample vial during displacement by helium. It looks like optimal currents are maintained until about 2000s, which corresponds to about 2mL of CO2. Stable ratios until about 1.25 mL CO2. Can we back calculate flow to source if current falls off below \~2ul/min CO2?
heflow <- 244 / 60 #flow in ul/s data.frame(x = 0:8000) %>% ggplot(aes(x)) + stat_function(fun=function(x) concCO2(x, r = heflow) * 7, aes(color = "CO2")) + scale_color_manual("Gas", values = c("green", "blue")) + labs(title = "CO2 in vial over time", subtitle = "7mL vial, 244uL/min helium", x = "Time (s)", y = "CO2 (mL)")
data.frame(x = 0:8000) %>% ggplot(aes(x)) + stat_function(fun=function(x) concCO2(x, r = heflow) * 30, aes(color = "CO2")) + scale_color_manual("Gas", values = c("green", "blue")) + labs(title = "CO2 flow over time", subtitle = "7mL vial, 244uL/min helium", x = "Time (s)", y = "CO2 (uL/s)")
Add flow at time to data and plot current vs flow. 30 ul/m is estimate of capillary flow based on past results.
data <- data %>% mutate(co2flow = concCO2(cum_acqtime, r = heflow) * 30) #30ul/min ggplot(data, aes(co2flow, he12C)) + geom_smooth(span = .3, se=FALSE) + geom_point() + labs(title = "Vial dillution current", subtitle = "250uL/min displacement, 30ul/min delivery", x = "CO2 flow (ul/m)", y = "12C current (uA)")
Ratio vs Current
data %>% filter(Pos == 5) %>% ggplot(aes(he12C, normFm)) + geom_point()
2021-04-09 Carbonate samples
Measured carbonate samples prepared on gasbench. Samples included TIRI-F, TIRI-I, and NOSAMS2. Normalized to gas standards or solid OX-I.
All samples measured on fresh targets after a 10 minute "burn in" to reduce blanks.
Using short capillaries from 2021-04-02. He flow set to 250ul/min.
Samples
carb_data <- tribble(~vial, ~name, ~rec_num, ~mass, 1, "NOSAMS2_1", 38809, 31.8, 2, "NOSAMS2_2", 38809, 32.3, 3, "NOSAMS2_3", 38809, 33.5, 4, "TIRI-I_1", 17185, 33.1, 5, "TIRI-I_2", 17185, 31.8, 6, "TIRI-I_3", 17185, 31.8, 7, "TIRI-F_1", 2138, 31.9, 8, "TIRI-F_2", 2138, 33.4, 9, "TIRI-F_3", 2138, 32.4 ) write_csv(carb_data, here("data/carb_data_2021-04-08.csv")) carb_data
Data
Normalizing to gas standards in pos 26 and 37.
data <- get_hgis_data(here("data/USAMS040121R.txt"), as.Date("2021-04-09")) %>% mutate(Num = ifelse(Pos %in% c(22, 24, 25, 27, 38), "U", Num)) stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat))
Currents and ratios vs time.
plot_hgis_time(data, normFm, ce*normFm) + ggtitle("Sample ratio vs. time")
plot_hgis_time(data, he12C) + ggtitle("Sample current vs. time")
Summary of normalized data per sample
std <- getStdTable() ds <- data %>% filter(!outlier) %>% sum_hgis() %>% mutate(rec_num = case_when(str_starts(Sample.Name, "LiveGas") ~ 101730, str_starts(Sample.Name, "C1") ~ 83028, str_starts(Sample.Name, "TIRI-F") ~ 2138, str_starts(Sample.Name, "TIRI-I") ~ 17185, str_starts(Sample.Name, "C2") ~ 1082, str_starts(Sample.Name, "NOSAMS2") ~ 38809, str_starts(Sample.Name, "DeadGas") ~ 72446)) %>% left_join(select(std, rec_num, fm_consensus), by = "rec_num") %>% mutate(fm_consensus = case_when(rec_num == 101730 ~ 1.0398, rec_num == 72446 ~ 0.0013, TRUE ~ fm_consensus)) ds[-(1:4),-4]
data %>% filter(as.numeric(Pos) > 5) %>% ggplot(aes(pos_name, normFm)) + geom_boxplot(fill = "slategray1") + theme(axis.text.x = element_text(angle = 45, hjust=1))
Agreement of replicates
Summary statistics to look at reproducibility, accuracy between targets/vials.
ds %>% filter(as.numeric(Pos) > 5) %>% mutate(Name = str_remove(Sample.Name, "_.$")) %>% group_by(Name, fm_consensus) %>% summarize(Fm.mean = mean(mean), Fm.sd = sd(mean), Cur.mean = mean(Cur), N = n())
Agreement of replicates is better than 2% for all samples. Samples have higher Fm than consensus. I think most of this is due to blanks, but the low gas standard also would push unknown Fm's up.
Blank correction
Testing correcting for blank using a mass-independent large blank correction.
Using TIRI-F as blank, naive with no adjustment for graphite value of blank. This pushes Fm in the right direction, but comparing samples of the same type, there's a lot of variability that seems to be tied to sample current.
fmblank <- ds %>% filter(str_starts(Sample.Name, "TIRI-F")) %>% pull(mean) %>% mean() fmblank ds <- ds %>% mutate(fm_lbc = doLBC(mean, fmblank, 1.0398), diff = fm_lbc - fm_consensus, normFm = normFm(fm_lbc, fm_consensus), sigma = sigma(fm_lbc, fm_consensus, merr)) ds %>% filter(as.numeric(Pos) > 4) %>% select(Sample.Name, Cur, fm_consensus, mean, fm_lbc, diff) ds %>% write_csv(here("data_analysed/USAMS040121.csv"))
ggplot(ds, aes(fm_consensus, diff)) + geom_hline(yintercept = 0) + geom_errorbar(aes(ymin = diff - merr, ymax = diff + merr)) + geom_point() + ggtitle("Difference in LBC-corrected vs consensus")
ggplot(ds, aes(fm_consensus, sigma)) + geom_hline(yintercept = 0) + geom_point() + ggtitle("Sigma of LBC-corrected vs consensus")
It looks like the normalization may be off a bit. Note positive linear correlation of Fm consensus difference with Fm of sample. Gas standards run at end of day were higher than those from the beginning. NOSAMS-2 was run early in the day, so closer to standards with lower ratios.
2021-04-16 Carbonate samples
Measured carbonate samples prepared on gasbench. Samples included C-1, C-2, and NOSAMS2. Normalized to gas standards or solid OX-I.
All samples measured on fresh targets after a 10 minute "burn in" to reduce blank.
Using short capillaries from 2021-04-02. He flow set to 250ul/min. Had to shorten delivery capillary by \~10cm at start of first C-2 due to acid clog. Needle blown out with compressed air.
Samples
carb_data <- tribble(~vial, ~name, ~rec_num, ~mass, 1, "C-1_1", 83028, 30.05, 2, "C-1_2", 83028, 36.00, 3, "C-1_3", 83028, 31.88, 4, "C-2_1", 1082, 32.57, 5, "C-2_2", 1082, 35.50, 6, "C-2_3", 1082, 34.10, 7, "NOSAMS-2_1", 38809, 35.20, 8, "NOSAMS-2_2", 38809, 32.19, 9, "NOSAMS-2_3", 38809, 33.20 ) write_csv(carb_data, here("data/carb_data_2021-04-15.csv")) carb_data
Data
Normalizing to gas standards in pos 25, 26, 36 and 37.
data <- get_hgis_data(here("data/USAMS041521R.txt"), as.Date("2021-04-16"))
Currents and ratios vs time.
data %>% filter(!(Pos %in% 21:24)) %>% plot_hgis_time(normFm, ce*normFm) + ggtitle("Sample ratio vs. time")
data %>% filter(!(Pos %in% 21:24)) %>% plot_hgis_time(he12C) ggtitle("Sample current vs time")
Summary of normalized data per sample
std <- getStdTable() ds <- data %>% filter(!outlier) %>% sum_hgis() %>% mutate(rec_num = case_when(str_starts(Sample.Name, "LiveGas") ~ 101730, str_starts(Sample.Name, "C-1") ~ 83028, str_starts(Sample.Name, "TIRI-F") ~ 2138, str_starts(Sample.Name, "TIRI-I") ~ 17185, str_starts(Sample.Name, "C-2") ~ 1082, str_starts(Sample.Name, "NOSAMS-2") ~ 38809, str_starts(Sample.Name, "DeadGas") ~ 72446)) %>% left_join(select(std, rec_num, fm_consensus), by = "rec_num") %>% mutate(fm_consensus = case_when(rec_num == 101730 ~ 1.0398, rec_num == 72446 ~ 0.0013, TRUE ~ fm_consensus)) ds[-(1:4),-3]
data %>% filter(as.numeric(Pos) > 4) %>% ggplot(aes(pos_name, normFm)) + geom_boxplot(fill = "slategray1") + theme(axis.text.x = element_text(angle = 45, hjust=1))
Agreement of replicates
Summary statistics to look at reproducibility, accuracy between targets/vials.
ds %>% filter(as.numeric(Pos) > 4) %>% mutate(Name = str_remove(Sample.Name, "_.$")) %>% group_by(Name, fm_consensus) %>% summarize(Fm.mean = mean(mean), Fm.sd = sd(mean), Cur.mean = mean(Cur), N = n())
Blank correction
Testing correcting for blank using mass-independent large blank correction.
Using C-1 with no adjustment for graphite value of blank.
fmblank <- ds %>% filter(str_starts(Sample.Name, "C-1")) %>% pull(mean) %>% mean() fmblank ds <- ds %>% mutate(fm_lbc = doLBC(mean, fmblank, 1.0398), diff = fm_lbc - fm_consensus, normFm = normFm(fm_lbc, fm_consensus), sigma = sigma(fm_lbc, fm_consensus, merr)) ds %>% filter(as.numeric(Pos) > 4) %>% select(Sample.Name, Cur, fm_consensus, mean, merr, fm_lbc, diff) ds %>% write_csv(here("data_analysed/USAMS041521.csv"))
ggplot(ds, aes(fm_consensus, diff)) + geom_hline(yintercept = 0) + geom_errorbar(aes(ymin = diff - merr, ymax = diff + merr)) + geom_point() + ggtitle("Difference in LBC-corrected vs consensus")
ggplot(ds, aes(fm_consensus, sigma)) + geom_hline(yintercept = 0) + geom_point() + ggtitle("Sigma of LBC-corrected vs consensus")
There may be some issues with normalization. The initial two standards on both of the last carbonate runs were lower than the two at the end of the run. On 4/9, first two standards were Fm 1.03 and 0.99, last two were 1.05 and 1.07. On 4/16, first two were 1.02 and 1.04, last two were both 1.05. NOSAMS2 were run first on 4/9 and were high after blank correction. On 4/16 they were run last and were low.
Work for SNEAP 2021 presentation
Combining data
A weak attempt at combining data. The solution is an analysis script to normalize, average runs, and blank correct by wheel.
ADD MOAR DATA
data<- get_hgis_data(here("data/USAMS101320R.txt"), as.Date("2020-11-17")) %>% mutate(Num = ifelse(Pos %in% c(2, 4), "S", ifelse(Num == "S", "U", Num))) stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data1013 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat)) data <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2020-12-04")) %>% mutate(Num = ifelse(Pos %in% c(2, 4), "S", ifelse(Num == "S", "U", Num))) stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data1204 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat)) data1211 <- get_hgis_data(here("data/USAMS120320R.txt")) %>% filter(Pos %in% 1:4 | as.Date(ts) == "2020-12-11") data <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2020-12-18")) %>% filter(Pos != 0) %>% mutate(Num = ifelse(Pos %in% c(2, 4), "S", ifelse(Num == "S", "U", Num))) stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data1218 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat)) data0108 <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2021-01-08")) # no solid standards # df <- get_hgis_data(here("data/USAMS120320R.txt"), as.Date("2021-01-15")) # stdrat <- mean(df$cor1412he[df$pos_name == "8 - LiveGasOS" & !df$outlier]) # data0115 <- mutate(df, normFm = norm_gas(cor1412he, stdrat)) data <- get_hgis_data(here("data/USAMS020521R.txt")) %>% mutate(Num = ifelse(Pos %in% c(2, 4), "S", ifelse(Num == "S", "U", Num))) stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data205 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat)) data <- get_hgis_data(here("data/USAMS030421R.txt"), as.Date("2021-03-05")) %>% mutate(Num = ifelse(Pos %in% c(2, 4), "S", ifelse(Num == "S", "U", Num))) stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data304 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat)) data <- get_hgis_data(here("data/USAMS040121R.txt"), as.Date("2021-04-09")) %>% mutate(Num = ifelse(Pos %in% c(22, 24), "S", ifelse(Num == "S", "U", Num))) stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data409 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat)) data <- get_hgis_data(here("data/USAMS041521R.txt"), as.Date("2021-04-16")) %>% mutate(Num = ifelse(Pos %in% c(22, 24), "S", ifelse(Num == "S", "U", Num))) stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data415 <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat)) std <- getStdTable() data <- rbind(data1013, data1204, data1211, data1218, data205, data304, data409, data415) %>% mutate(rec_num = case_when(str_detect(Sample.Name, "LiveGas") ~ 101730, str_starts(Sample.Name, "C-1") ~ 83028, str_starts(Sample.Name, "C1") ~ 83028, str_starts(Sample.Name, "TIRI-F") ~ 2138, str_starts(Sample.Name, "TIRI-I") ~ 17185, str_starts(Sample.Name, "C-2") ~ 1082, str_starts(Sample.Name, "C2") ~ 1082, str_starts(Sample.Name, "NOSAMS") ~ 38809, str_detect(Sample.Name, "DeadGas") ~ 72446)) %>% left_join(select(std, rec_num, fm_consensus), by = "rec_num") %>% mutate(fm_consensus = case_when(rec_num == 101730 ~ 1.0398, rec_num == 72446 ~ 0.0013, TRUE ~ fm_consensus)) data %>% distinct(wheel, Pos, Sample.Name, fm_consensus, rec_num) %>% filter(fm_consensus > 0.9)
Current dependent blank with combined data
unique(data$Sample.Name)
Modern
data %>% filter(!(he12C < 2.5 & normFm > 1.05) & fm_consensus > 0.9) %>% #filter(fm_consensus > 0.9) %>% #filter(str_detect(Sample.Name, "LiveGas")) %>% ggplot(aes(he12C, normFm)) + #geom_smooth(method = "lm") + geom_point() data %>% select(wheel, Pos, Sample.Name, he12C, fm_consensus, normFm) %>% filter(he12C < 2.5 & normFm > 1.05, fm_consensus > 0.9)
Dead
data %>% #filter(str_detect(Sample.Name, "DeadGas")) %>% filter(fm_consensus < 0.1 & normFm < 0.3 & Sample.Name != "DilDeadGas6") %>% ggplot(aes(he12C, normFm)) + geom_point() data %>% select(wheel, Pos, Sample.Name, he12C, fm_consensus, normFm) %>% filter(he12C < 1 & normFm < .1, fm_consensus < 0.3) filter(data, Sample.Name == "DilDeadGas6")
Summarize by sample
ds <- sum_hgis(data) %>% filter((str_detect(Sample.Name, "LiveGas") & Cur > 1) | (str_detect(Sample.Name, "DeadGas") & Cur > 1) | str_detect(Sample.Name, "Blank")) %>% mutate(Gas = case_when(str_detect(Sample.Name, "LiveGas") ~ "LiveGas", str_detect(Sample.Name, "DeadGas") ~ "DeadGas", TRUE ~ "Blank"), Cur_inv = 1/Cur)
Modern
live <- ds %>% filter(str_detect(Sample.Name, "LiveGas") & Cur > 1) %>% ggplot(aes(Cur, mean)) + # geom_smooth(method = "lm") + geom_pointrange(aes(ymin = mean - merr, ymax = mean + merr)) + geom_point()
Dead
dead <- ds %>% filter(str_detect(Sample.Name, "DeadGas")) %>% ggplot(aes(Cur, mean)) + #geom_smooth(method = "lm") + geom_pointrange(aes(ymin = mean - merr, ymax = mean + merr)) + geom_point() filter(ds, Cur < 1)
live / dead
Modelling current dependent blank
Fit linear models
fits <- ds %>% ungroup() %>% nest(data = -Gas) %>% mutate(fit = map(data, ~lm(mean ~ Cur_inv, data = .x)), tidied = map(fit, tidy)) %>% unnest(tidied) blankfit <- fits %>% select(Gas, term, estimate) %>% pivot_wider(names_from = c(Gas, term), values_from = estimate) %>% mutate(inv_m_blank = -(`DeadGas_(Intercept)` - `LiveGas_(Intercept)`)/(DeadGas_Cur_inv - LiveGas_Cur_inv), Fm_blank = `DeadGas_(Intercept)` + DeadGas_Cur_inv * inv_m_blank, m_blank = 1/inv_m_blank) blankfit %>% select(Fm_blank, m_blank) blankfit
Fm vs. inverse mass with fits.
ggplot(ds, aes(Cur_inv, mean, color = Gas)) + geom_abline(slope = blankfit$LiveGas_Cur_inv, intercept = blankfit$`LiveGas_(Intercept)`) + geom_abline(slope = blankfit$DeadGas_Cur_inv, intercept = blankfit$`DeadGas_(Intercept)`) + geom_pointrange(aes(ymin = mean - merr, ymax = mean + merr), size = .3) + xlim(0, 10) + #ylim(0, 1.2) ggtitle("Fm of live and dead CO2 vs. inverse 12C current", subtitle = "Fit lines should cross at Fm and current of blank")
Gas standards
Use above data with gas standards normalized to solid OX-I
rownames(ds) ds %>% filter(Gas == "LiveGas") %>% ggplot(aes(x = as.factor(as.numeric(rownames(.))), y = mean)) + geom_pointrange(aes(ymin = mean - merr, ymax = mean + merr))
2021-05-12 Vial dillution tests
Data
data <- get_hgis_data(here("data/USAMS041521R.txt"), as.Date("2021-05-12"))
Currents and ratios vs time.
data %>% plot_hgis_time(cor1412he, ce*cor1412he) + ggtitle("Sample ratio vs. time")
data %>% plot_hgis_time(he12C) + ggtitle("Sample current vs. time")
2021-05-14 Carbonate samples
Samples
carb_data <- tribble(~vial, ~name, ~rec_num, ~mass, 1, "C-1_1", 83028, 31.61, 2, "C-1_2", 83028, 34.30, 2, "C-1_3", 83028, 33.17, 4, "TIRI-I_1", 17185, 32.74, 5, "TIRI-I_2", 17185, 31.58, 6, "TIRI-I_3", 17185, 33.25, 7, "NOSAMS-2_1", 38809, 33.31, 8, "NOSAMS-2_2", 38809, 31.89, 9, "NOSAMS-2_3", 38809, 36.47 ) write_csv(carb_data, here("data/carb_data_2021-05-14.csv")) carb_data
Data
Normalizing to gas standards in pos 25, 26, 36 and 37.
data <- get_hgis_data(here("data/USAMS051421R.txt"), as.Date("2021-05-14")) %>% mutate(Num = ifelse(Pos %in% c(15, 65), "U", Num)) stdrat <- mean(data$cor1412he[data$Num == "S" & !data$outlier]) data <- data %>% mutate(normFm = norm_gas(cor1412he, stdrat))
Currents and ratios vs time.
data %>% filter(!(Pos %in% 1:4)) %>% plot_hgis_time(normFm, ce*normFm) + ggtitle("Sample ratio vs. time")
data %>% filter(!(Pos %in% 1:4)) %>% plot_hgis_time(he12C) ggtitle("Sample current vs time")
Summary of normalized data per sample
std <- getStdTable() ds <- data %>% filter(!outlier) %>% sum_hgis() %>% mutate(rec_num = case_when(str_starts(Sample.Name, "LiveGas") ~ 101730, str_starts(Sample.Name, "C1") ~ 83028, str_starts(Sample.Name, "TIRI-F") ~ 2138, str_starts(Sample.Name, "TIRI-I") ~ 17185, str_starts(Sample.Name, "C-2") ~ 1082, str_starts(Sample.Name, "NOSAMS2") ~ 38809, str_starts(Sample.Name, "DeadGas") ~ 72446)) %>% left_join(select(std, rec_num, fm_consensus), by = "rec_num") %>% mutate(fm_consensus = case_when(rec_num == 101730 ~ 1.0398, rec_num == 72446 ~ 0.0013, TRUE ~ fm_consensus)) ds[-(1:4),]
data %>% filter(as.numeric(Pos) > 4) %>% ggplot(aes(pos_name, normFm)) + geom_boxplot(fill = "slategray1") + theme(axis.text.x = element_text(angle = 45, hjust=1))
Agreement of replicates
Summary statistics to look at reproducibility, accuracy between targets/vials.
ds %>% filter(as.numeric(Pos) > 4) %>% mutate(Name = str_remove(Sample.Name, "_.$")) %>% group_by(Name, fm_consensus) %>% summarize(Fm.mean = mean(mean), Fm.sd = sd(mean), Cur.mean = mean(Cur), N = n())
Blank correction
Testing correcting for blank using mass-independent large blank correction.
Using C-1 with no adjustment for graphite value of blank.
fmblank <- ds %>% filter(str_starts(Sample.Name, "C1")) %>% pull(mean) %>% mean() fmblank ds <- ds %>% mutate(fm_lbc = doLBC(mean, fmblank, 1.0398), diff = fm_lbc - fm_consensus, normFm = normFm(fm_lbc, fm_consensus), sigma = sigma(fm_lbc, fm_consensus, merr)) ds %>% filter(as.numeric(Pos) > 4) %>% select(Sample.Name, Cur, fm_consensus, mean, merr, fm_lbc, diff) ds %>% write_csv(here("data_analysed/USAMS051421.csv"))
ggplot(ds, aes(fm_consensus, diff)) + geom_hline(yintercept = 0) + geom_errorbar(aes(ymin = diff - merr, ymax = diff + merr)) + geom_point() + ggtitle("Difference in LBC-corrected vs consensus")
ggplot(ds, aes(fm_consensus, sigma)) + geom_hline(yintercept = 0) + geom_point() + ggtitle("Sigma of LBC-corrected vs consensus")
2021-05-20 Dillution tests
Data
data <- get_hgis_data(here("data/USAMS051421R.txt"), as.Date("2021-05-20"))
Currents and ratios vs time.
data %>% plot_hgis_time(cor1412he, ce*cor1412he) + ggtitle("Sample ratio vs. time")
data %>% plot_hgis_time(he12C) + ggtitle("Sample current vs. time")
2021-05-28 Pendleton samples
First run of Simon Pendleton's samples, along with C-1, C-2 and modern standard. Reworked data reduction and did analysis in separate file.
2021-06-04 Small(er) sample tests
Tested 5, 10, and 15 mg of C-2 carbonate in vials backfilled with Helium.
Method
- Weigh sample into vial
- Evacuate and acidify as normal with gilson
- Backfill to 1 ATM with He-filled syringe 2h after acidification.
Allowed vacuum in vial to pull in He from syringe, then added 1 additional mL He. Should calculate volume CO2 in vial and measure He added to check that things make sense.
Used same capillaries as usual, but decreased He displacement flow to 140 ul/min. Source capillary flow assumed to be around 30ul/min.
Samples
carb_data <- tribble(~number, ~name, ~rec_num, ~mass, 1, "C-2_1", 1082, 6.03, 2, "C-2_2", 1082, 11.10, 3, "C-2_3", 1082, 15.52 ) write_csv(carb_data, here("data/carb_data_2021-06-04.csv")) carb_data
Data
df <- process_hgis_results(here("data/USAMS052621R.txt"), as.Date("2021-06-04")) #, standards)
Currents and ratios vs time.
df %>% filter(!(pos %in% 1:4)) %>% plot_hgis_time(norm_ratio, sig_norm_ratio) + ggtitle("Sample ratio vs. time")
df %>% filter(!(pos %in% 1:4)) %>% plot_hgis_time(he12C) + ggtitle("Sample current vs time")
Results
Note: no blanks run, so need to use average blank if LBC needed.
df_sum <- sum_hgis_targets(df) %>% norm_hgis()
df_sum %>% filter(as.numeric(pos) > 4) %>% select(pos, sample_name, he12C, n_runs, norm_ratio, sig_norm_ratio) %>% arrange(sample_name)
df_sum %>% filter(as.numeric(pos) > 4) %>% ggplot(aes(sample_name, norm_ratio)) + geom_pointrange(aes(ymin = norm_ratio - sig_norm_ratio, ymax = norm_ratio + sig_norm_ratio), size = .5) + theme(axis.text.x = element_text(angle = 45, hjust=1)) + labs(x = NULL, y = "Fraction Modern")
Agreement of replicates
Compare results for samples with more than one replicate.
df_sum %>% filter(as.numeric(pos) > 4) %>% compare_replicates()
Agreement with consensus
df_sum %>% filter(as.numeric(pos) > 4, !is.na(fm_consensus)) %>% select(sample_name, fm_consensus, norm_ratio, sig_norm_ratio) %>% mutate(Fm_diff = norm_ratio - fm_consensus, sigma = amstools::sigma(norm_ratio, fm_consensus, sig_norm_ratio)) %>% arrange(sample_name)
2021-06-11 Small(er) sample tests
Tested 8mg of C-1, C-2, and NOSAMS2 carbonate in vials backfilled with Helium.
Method
- Weigh sample into vial
- Evacuate and acidify as normal with gilson
- Backfill to 1 ATM with He-filled syringe 2h after acidification.
Allowed vacuum in vial to pull in He from syringe, then added 1 additional mL He. Should calculate volume CO2 in vial and measure He added to check that things make sense.
Used same capillaries as usual, but decreased He displacement flow to 140 ul/min. Source capillary flow assumed to be around 30ul/min.
Samples
carb_data <- tribble(~name, ~rec_num, ~carbonate_mass, ~helium_added, "C-1_1", 83028, 8.29, 5, "C-1_2", 83028, 9.01, 5, "C-1_3", 83028, 9.00, 5, "C-2_1", 1082, 8.23, 5, "C-2_2", 1082, 8.77, 5, "C-2_3", 1082, 7.34, 5, "NOSAMS-2_1", 38809, 7.72, 5, "NOSAMS-2_2", 38809, 7.69, 5, "NOSAMS-2_3", 38809, 8.86, 5 ) write_csv(carb_data, here("data/carb_data_2021-06-04.csv")) #insert_samples(carb_data) carb_data
Data
df <- process_hgis_results(here("data/USAMS061121R.txt"), as.Date("2021-06-11")) #, standards) #insert_raw(df)
Currents and ratios vs time.
df %>% filter(pos > 2) %>% plot_hgis_time(norm_ratio, sig_norm_ratio, outlier = !.$ok_calc) + ggtitle("Sample ratio vs. time")
df %>% filter(pos > 2) %>% plot_hgis_time(he12C, outlier = !.$ok_calc) + ggtitle("Sample current vs time")
Results
Note: no blanks run, so need to use average blank if LBC needed.
df_sum <- sum_hgis_targets(df) %>% norm_hgis() %>% blank_cor_hgis() #insert_results(df_sum)
df_sum %>% filter(pos > 2) %>% mutate(he12C = he12C*1E6) %>% select(Positon = pos, `Sample Name` = sample_name, `12C Current (uA)` = he12C, n_runs, fm_corr, sig_fm_corr) %>% arrange(`Sample Name`)
df_sum %>% plot_hgis_summary()
Agreement of replicates
Compare results for samples with more than one replicate.
df_sum %>% filter(pos > 2) %>% compare_replicates()
Agreement with consensus
cons <- df_sum %>% filter(pos > 2, !is.na(fm_consensus)) %>% select(sample_name, fm_consensus, fm_corr, sig_fm_corr) %>% mutate(fm_diff = fm_corr - fm_consensus, sigma = amstools::sigma(fm_corr, fm_consensus, sig_fm_corr)) %>% arrange(sample_name) cons knitr::kable(cons)
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.