Ca due to reductive cleaning and on sample loss during cleaning

Chemical Geology 420 (2016) 23–36

Contents lists available at ScienceDirect

Chemical Geology journal homepage: www.elsevier.com/locate/chemgeo

Effect of preservation state of planktonic foraminifera tests on the decrease in Mg/Ca due to reductive cleaning and on sample loss during cleaning Heather J.H. Johnstone a,⁎, William Lee b, Michael Schulz a a b

MARUM — Center for Marine Environmental Sciences, University of Bremen, Germany MACSI, Department of Mathematics and Statistics, University of Limerick, Limerick, Ireland

a r t i c l e

i n f o

Article history: Received 30 April 2015 Received in revised form 18 September 2015 Accepted 31 October 2015 Available online 2 November 2015 Keywords: Foraminifera Mg/Ca Reductive cleaning Calcite preservation Cd-cleaning Mg-cleaning

a b s t r a c t Four species of planktic foraminifera from core-tops spanning a depth transect on the Ontong Java Plateau were prepared for Mg/Ca analysis both with (Cd-cleaning) and without (Mg-cleaning) a reductive cleaning step. Reductive cleaning caused etching of foraminiferal calcite, focused on Mg-rich inner calcite, even on tests which had already been partially dissolved at the seafloor. Despite corrosion, there was no difference in Mg/Ca of Pulleniatina obliquiloculata between cleaning methods. Reductive cleaning decreased Mg/Ca by an average (all depths) of ~4% for Globigerinoides ruber white and ~10% for Neogloboquadrina dutertrei. Mg/Ca of Globigerinoides sacculifer (above the calcite saturation horizon only) was 5% lower after reductive cleaning. The decrease in Mg/ Ca due to reductive cleaning appeared insensitive to preservation state for G. ruber, N. dutertrei and P. obliquiloculata. Mg/Ca of Cd-cleaned G. sacculifer appeared less sensitive to dissolution than that of Mgcleaned. Mg-cleaning is adequate, but SEM and contaminants (Al/Ca, Fe/Ca and Mn/Ca) show that Cd-cleaning is more effective for porous species. A second aspect of the study addressed sample loss during cleaning. Lower yield after Cd-cleaning for G. ruber, G. sacculifer and N. dutertrei confirmed this to be the more aggressive method. Strongest correlations between yield and Δ[CO2− 3 ] in core-top samples were for Cd-cleaned G. ruber (r = 0.88, p = 0.020) and Cd-cleaned P. obliquiloculata (r = 0.68, p = 0.030). In a down-core record (WIND28K) correlation, r, between yield values N30% and dissolution index, XDX, was − 0.61 (p = 0.002). Where cleaning yield b 30% most Mg-cleaned Mg/Ca values were biased by dissolution. © 2015 Elsevier B.V. All rights reserved.

1. Introduction The seawater temperature proxy based on the ratio of Mg to Ca in foraminiferal calcite was developed during the 1990s (Nürnberg, 1995; Nürnberg et al., 1996; Rosenthal et al., 1997; Hastings et al., 1998; Mashiotta et al., 1999; Lea et al., 1999; Elderfield and Ganssen, 2000) and is now a routine part of paleoceanography. Thorough cleaning of the foraminifera tests is necessary before analysis to remove contaminants which would otherwise bias trace-metal concentrations. Methods in current use derive from the protocol of Boyle (1981). Cleaning involves first breaking open the test chambers, fragments are then rinsed several times with water. This removes fine clays and any other sedimentary material trapped inside the test. An oxidative cleaning step (hydrogen peroxide buffered with sodium hydroxide) is used to remove organic material. A reductive cleaning step (anhydrous hydrazine–ammonium hydroxide–ammonium citrate solution) removes coatings of metal oxides. Test fragments undergo further rinsing with water ⁎ Corresponding author. E-mail addresses: [email protected] (H.J.H. Johnstone), [email protected] (W. Lee), [email protected] (M. Schulz).

http://dx.doi.org/10.1016/j.chemgeo.2015.10.045 0009-2541/© 2015 Elsevier B.V. All rights reserved.

followed by a weak acid leach before the sample is finally dissolved for analysis. When Mg/Ca is the main interest, a simplified method which excludes the reductive step is often used. This method is referred to as “Mg-cleaning”, while the method including a reductive step is called “Cd-cleaning”. Contaminant phases which contain Mg, such as marine clays and some metal oxide coatings (Pena et al., 2005; Weldeab et al., 2006), are associated with Fe, Al and Mn and these elements are often monitored for quality control. A number of studies find that Cd-cleaning gives somewhat lower Mg/Ca values than Mg-cleaning, even where Fe and Mn values suggest that there was no high Mg contaminant phase to be removed. The interlaboratory comparison study of Rosenthal et al. (2004) used several species of planktonic foraminifera (Globigerinoides ruber white, Globigerinoides sacculifer, Pulleniatina obliquiloculata, Globigerina bulloides, Orbulina universa) and concluded that Mg/Ca was ~ 15% lower when Cd-cleaning rather than Mg-cleaning was used. This is in agreement with Barker et al. (2003) who found that Mg/Ca after Cdcleaning could be 10–15% lower than with Mg-cleaning. A value of 10%, or a fixed offset of 0.2 mmol/mol is often used for conversion between the two methods, so that datasets cleaned using different methods can later be combined (e.g. Elderfield et al., 2006; Sadekov

24

H.J.H. Johnstone et al. / Chemical Geology 420 (2016) 23–36

et al., 2010). These amounts are potentially significant in terms of calculated temperature. For many calibrations between Mg/Ca and temperature (e.g. Anand et al., 2003), 10% is equivalent to ~1 °C of temperature. This is a large amount considering that the entire glacial–interglacial change in sea-surface temperature in the tropics is ~ 3 °C (e.g. Lea, 2004). For benthic species 0.2 mmol/mol could be even more than 10% of the Mg/Ca value, and so represent N1 °C. Conversion values must be selected with care, as a number of studies show species specific and/or study specific, differences in the offset between the two cleaning methods. Martin and Lea (2002) found that reductive cleaning made no difference to the Mg/Ca of (poorly preserved) Uvigerina species, but lead to a 10% decrease in the Mg/Ca of (well preserved) Cibicidoides wuellerstorfi. Barker et al. (2003) confirmed no offset in Uvigerina but Yu et al. (2007) found a 10% decrease in Mg/Ca with Cd-cleaning for three species of benthic foraminifera including Uvigerina species and C. wuellerstorfi. Barker et al. (2003) found large offsets (N15%) for G. ruber white, Globorotalia hirsuta and Neogloboquadrina pachyderma, ~ 10% for Globorotalia truncatulinoides and Globorotalia inflata, and no offset for G. sacculifer or G. bulloides. Studies using Laser Ablation (LA), rather than bulk solution, also found no difference in Mg/Ca after reductive cleaning for G. bulloides (Marr et al., 2013) and O. universa (Vetter et al., 2013). Xu et al. (2008) found a difference of 6.5% for G. ruber white and 1.8% for P. obliquiloculata. Regenberg et al. (2014) found an offset of 3% for a group of six planktonic species including G. ruber white (also G. sacculifer, P. obliquiloculata, Neogloboquadrina dutertrei, Globorotalia G. menardii and G. truncatulinoides) which were affected by dissolution. Mg-cleaning has the advantage that the method requires less time. It also avoids the use of the toxic reagent hydrazine and leads to less loss of sample during cleaning; however, there is no doubt that Cd-cleaning removes a wider range of contaminants (Boyle, 1981; Weldeab et al., 2006; Pena et al., 2005). For these reasons, both methods are likely to remain in common use and it is important to understand what controls the offset in Mg/Ca. One explanation for the decrease in Mg/Ca after Cd-cleaning is that the reagent hydrazine dissolves test calcite slightly. Dissolution of planktonic foraminifera at the sea floor results in lower Mg/Ca, attributed to ‘selective dissolution’, as calcite containing Mg is more soluble than pure calcite (Lorens et al., 1977; Russell et al., 1994; Brown and Elderfield, 1996; Hastings et al., 1998; Rosenthal et al., 2000). It has been suggested, therefore, that lower Mg/Ca after Cd-cleaning is caused by dissolution of test calcite by the reagents used in reductive cleaning (Barker et al., 2003; Rosenthal et al., 2004). Yu et al. (2007) established that the citrate added to buffer the reducing agent (hydrous hydrazine), when applied in isolation of the reducing agent, can dissolve foraminiferal calcite. Barker et al. (2003) raised the question of whether the offset in Mg/ Ca between cleaning methods is sensitive to preservation state of tests. The question has not been specifically addressed. Although Regenberg et al. (2014) compared splits of samples (cleaned either by standard Mg-cleaning or by Mg-cleaning with a hydrazine rinse) with a range of preservation states, their data came from a wide geographical area (5°N to 20°N in the South China Sea) so there is a variation in initial values, and they did not consider the issue directly. Rosenthal et al. (2004) included G. ruber and G. sacculifer from a deep site (with poor calcite preservation) in their cleaning comparison study, but the two cleaning methods were carried out in different labs, so they are not directly comparable. Bian and Martin (2010) used N. dutertrei with a range of preservation states, and compared protocols with different amounts of citrate in the reductive cleaning reagent, but did not compare to standard Mg-cleaning. Barker et al. (2003) suggested that the magnitude of the offset between methods may be less for foraminifera from partially dissolved sediments. This would be the case if the decrease in Mg/Ca caused by reductive cleaning was sensitive to Mg concentration such that the offset in Mg/Ca between the two methods would be greater for wellpreserved, high Mg/Ca, tests than for partially dissolved tests with low

initial Mg/Ca. In this scenario, reductive cleaning would give a more reproducible result, insensitive to slight prior dissolution of the test, although less representative of initial Mg/Ca. Alternatively, partial dissolution of the test at the sea floor could allow greater penetration of corrosive reagents during cleaning, meaning that reductive cleaning would cause more dissolution, and associated leaching of Mg, in poorly preserved tests. The primary question of this study is: (1) Does preservation state of a sample control the decrease in Mg/ Ca caused by reductive cleaning? The second part of this study deals with cleaning, or analytical, “yield”, the proportion of the original sample which is actually analysed. Cleaning has a high attrition rate, and yield can be a small fraction of the original. There is anecdotal evidence that more material is lost from samples which are poorly preserved. Yield may be as much as 80% for wellpreserved tests cleaned by the Mg method but can fall to less than 10% for poorly preserved tests exposed to the more rigorous Cd-cleaning (M. Greaves, pers. com.). The effect of dissolution on Mg/Ca is a major drawback of this proxy. Tachikawa et al. (2008) suggest that yield may be worth monitoring as a dissolution indicator and assumed that low yield (b20%) meant that Mg/Ca was biased by supralysoclinal dissolution. However calibrations comparing yield to deep water calcite saturation are so far lacking. We provide these in order to address the question: (2) Does cleaning yield of a sample indicate the preservation state of foraminifera tests? 2. Material and methods In order to address the two questions posed in the Introduction we compare the effect of two cleaning methods, Mg-cleaning and Cdcleaning, on the Mg/Ca of four species of planktonic foraminifera from a depth transect on the Ontong Java Plateau (OJP) (McCorkle and Keigwin, 1994). The Mg-cleaned data presented here is a subset of that of Johnstone et al. (2011). Here we present (previously unpublished) Cd-cleaned Mg/Ca from the same sites. Prior to crushing and cleaning, the samples had been scanned by CT (computed tomography) for previous studies (Johnstone et al., 2010; Johnstone et al., 2011). CT is a non-destructive technique and did not affect element to Ca ratio of the scanned samples (Johnstone et al., 2011). Four species of planktonic foraminifera were used: G. ruber (white), G. sacculifer (with no sac-like final chamber), N. dutertrei and P. obliquiloculata. The four species are different physically. G. ruber and G. sacculifer have an open porous texture with surface ridges but little or no outer crust. N. dutertrei also has pores, but also has a thick outer crust. P. obliquiloculata has a smooth veneer on top of its calcite crust and no obvious pores (Hemleben et al., 1989). Samples span a range of deep water calcite saturation states 2− (Δ[CO2− 3 ]) from +13 to −20 μmol/kg. Δ[CO3 ] is defined as the differ2− 2− ence between [CO2− ] (measured [CO 3 IN SITU 3 ] at the site) and [CO3 ] 2− 2− SATURATION (calculated [CO3 ] value at saturation at the site). Δ[CO3 ]

Table 1 Details of cores used in this study. Core

1BC3 1.5BC33 2BC13 2.5BC37 3BC16 3BC24 4BC51 4.5BC53 5BC54 5.5BC58 6BC66

Lat.

Long.

Water depth

Δ[CO2− 3 ]

[° N]

[° W]

[m]

[μmol/kg]

−2.24 −1.00 −0.01 0.00 0.01 0.01 −0.02 −0.01 −0.01 0.00 0.00

−157.00 −157.85 −158.91 −159.48 −160.45 −160.43 −161.02 −161.39 −161.77 −162.22 −162.70

1616 2015 2301 2445 2959 2965 3411 3711 4025 4341 4400

13.8 9.3 4.9 4.3 −2.2 −2.3 −5.8 −11.8 −14.7 −22.4 −23.0

Calcite saturation (Δ[CO2− 3 ]) from Johnstone et al. (2010).

H.J.H. Johnstone et al. / Chemical Geology 420 (2016) 23–36

25

Table 2 Yield and Mg/Ca, Mn/Ca, Al/Ca, Fe/Ca for four species of planktic foraminifera cleaned using Mg-cleaning and Cd-cleaning methods. “Mg-c le a n in g ” m e th o d *

C o re

Nu. te s ts

T o ta l mass s a m p le [µ µg ]

Test mass (a ve ) [µ µg ]

Y ie ld

Mg /C a

[%]

S r/C a

“C d-c le a n in g ” m e th o d

A l/C a

Mn /C a

F e /C a

[m m o l [m m o l [µm o l /m o l] /m o l] /m o l]

[µm o l /m o l]

[m m o l /m o l]

Nu. te s ts

T o ta l mass s a m p le [µ µg ]

Test mass (a ve ) [µ µg ]

Y ie ld

Mg /C a

[%]

S r/C a

A l/C a

Mn /C a

F e /C a

µm o l [m m o l [m m o l [µ /m o l] /m o l] /m o l]

[µ µm o l /m o l]

[m m o l /m o l]

G. ruber (w hite ), 300 – 355 µ m 1B C 3 1.5B C 33 2B C 13 2.5B C 37 3B C 16 3B C 24

35 32 34 35 28 32

602 550 464 462 346 391

17 17 14 13 12 12

Average

38 34 26 32 20 40

4.81 4.47 4.13 4.24 4.18 4.28

1.51 1.47 1.43 1.47 1.46 1.49

92 22 26 57 22 19

4.6 3.6 5 7.4 2.3 3.3

29 20 44 52 0

32

4.35

1.47

40

4.4

29

31

565 487 448 491 415 348

32 32 34 34 34 31

18 15 13 14 12 11

Average of difference between pairs (Mg-cleaned minus Cd-cleaned)

33 23 24 7 2 10

4.49 4.51 4.13 4.07 4.17 3.65

1.49 1.48 1.47 1.49 1.45 1.45

13 17 6 21 n 6

1.2 0.5 0.3 1.2 n 0.8

12 0 0 0 0 0

17

4.17

1.47

13

0.8

2

15

0.18

0.00

31

4.0

27

42 36 29 33 27 35 1

3.90 3.86 3.58 3.70 3.58 3.64 3.54

1.38 1.36 1.35 1.35 1.34 1.35 1.35

16 22 4 17 16 4 n

-0.3 0.2 -0.3 0.5 0.0 -0.3 n

12 0 0 0 0 0 0

29

3.68

1.36

13

0.0

2

17

0.10

0.00

4

4.1

25

•

G. sacculifer (w ith o u t s a c ), 300 – 355 µ m 1B C 3 1.5B C 33 2B C 13 2.5B C 37 3B C 16 3B C 24 4B C 51

32 35 34 29 32 33 30

734 775 612 514 524 565 471

23 22 18 18 16 17 16

Average

58 51 57 47 25 52 35

4.02 4.24 3.76 3.85 3.41 3.54 3.64

1.41 1.38 1.34 1.36 1.34 1.33 1.34

27 32 6 11 18 8 21

2.8 2.5 0.9 12.3 2 3.6 1.3

50 36 23 12 0

46

3.78

1.36

18

3.6

27

49 19

34 34 36 40 40 40 40

767 710 642 754 633 633 689

23 21 18 19 16 16 17

Average of difference between pairs (Mg-cleaned minus Cd-cleaned)

•

N. dutertrei, 300 – 355 µ m 1B C 3 1.5B C 33 2B C 13 2.5B C 37 3B C 16 3B C 24 4B C 51 4.5B C 53 5B C 54 5.5B C 58 6B C 66

29 32 29 32 30 32 31 30 21 35 Not run

730 720 679 633 548 684 535 483 388 525

25 23 23 20 19 21 17 16 18 15

Average

42 49 50 44 35 38 24 16 50 12

2.28 2.10 1.82 1.97 1.74 1.75 1.33 1.44 1.40 1.08

1.36 1.36 1.36 1.36 1.35 1.36 1.38 1.39 1.42 1.44

80 25 10 16 12 5 36 14 19 18

5.3 4.1 4.5 3.4 1.5 1.4 1.1 0.5 1.1 0.6

31 28 16 15 0 0 0 0 20 0.000

36

1.69

1.38

24

2.4

11

37 39 42 40 40 37 42 47 Not run 44 30

899 891 992 867 693 766 680 710

24 23 24 22 17 21 16 15

25 48 32 27 19 20 19 22

2.13 1.93 1.76 1.73 1.42 1.42 1.19 1.24

1.39 1.38 1.36 1.36 1.36 1.37 1.39 1.39

19 17 3 7 1 -2 1 13

1.6 1.3 0.6 0.9 0.0 -0.3 -0.2 0.0

11 13 0 0 0 0 0 0

713 494

16 16

15 22

1.18 1.07

1.42 1.43

15 4

0.1 0.2-

0 0

25

1.51

1.39

8

0.4

2

9

0.17

-0.01

16

2.0

7

31 29 27 26 25 23 22 21

61 54 22 33 30 31 37 27

2.87 2.52 2.52 2.21 2.14 2.16 1.82 1.73

1.39 1.41 1.42 1.39 1.37 1.37 1.36 1.37

43 23 3 2 5 6 18 28

2.6 2.2 2.3 2.1 2.5 3.2 2.1 2.0

11 14 9 0

17 16

24 26

1.47 1.44

20 24

1.7 1.8

11

35

2.09

1.30 1.37 1.37

17

2.3

6

-11

0.01

0.03

0

2.6

3

Average of difference between pairs (Mg-cleaned minus Cd-cleaned) P. obliquiloculata, 300 – 355 µ m 1B C 3 1.5B C 33 2B C 13 2.5B C 37 3B C 16 3B C 24 4B C 51 4.5B C 53 5B C 54 5.5B C 58 6B C 66 Average

35 30 30 30 36 30 29 30 27 28

1147 792 817 813 not run 840 681 566 595 447 460

33 26 27 27

56 41 44 0

2.84 2.66 2.33 n

1.43 1.42 1.44 n

39 12 9 n

7.1 4.2 4.4 n

32 21 12 n

33 11 13 48 3 19

2.39 1.90 1.65 1.62 1.46 1.40

1.43 1.35 1.35 1.39 1.38 1.40

6 11 36 12 n 32

12.8 2.5 3 2.6 1.7 2.7

9 0 0

38 38 38 40 39 41 41 45

23 23 20 20 17 16

10 0 0

47 52

2.03

1.40

20

4.6

9

27

•

Average of difference between pairs (Mg-cleaned minus Cd-cleaned)

Element analysis by ICP-MS, except for Fe which was by ICP-OES. *Mg/Ca data first published in Johnstone et al. (2011). •Sample too small for ICP-MS, data shown is from ICP-OES; n: no data. 0 below calibration limit, value of 0 used. Non-significant average differences have grey background. Light grey background marks the 4 samples which had no partner.

1181 1115 1007 1039 979 947 922 925 Not run 797 822

6 0 0 6 0

26

H.J.H. Johnstone et al. / Chemical Geology 420 (2016) 23–36

values in Table 1 are from Johnstone et al. (2010). Previous studies on samples from the OJP show that they cover a range of preservation states, from well preserved to severely corroded. The effect of seafloor dissolution on the tests has been described (Johnstone et al., 2010). SEM showed the test calcite to be pitted and etched by corrosive waters. Corrosion was particularly apparent on the inner calcite of the test, which became cracked and flakey. Computed tomography (CT) (Johnstone et al., 2010) confirmed that the inner test calcite is particularly susceptible to dissolution; tests from increasingly undersaturated sites show progressive loss of inner chamber walls. In samples from the deepest OJP sites, only the outer crusts of N. dutertrei and P. obliquiloculata remain. The effect of dissolution was also shown in the mass and Mg/Ca of samples; both properties decreased as water depth increased. For the samples used in the current study, test mass in the deepest sample used is 29% lighter than that of the shallowest (1BC3 at 1616 m) for G. ruber, 32% for G. sacculifer, 40% for N. dutertrei, and 50% for P. obliquiloculata ((Johnstone et al., 2010), Table 2). Mg/Ca (Mg-cleaned) of the deepest sample of G. ruber is 11% lower than that of the shallowest. It is 9% lower for G. sacculifer, 51% for N. dutertrei and 53% for P. obliquiloculata ((Johnstone et al., 2011), Table 2). Samples were weighed before being crushed. This was used as initial mass in the calculation of analytical yield (this assumes that the weight of contaminant phases are negligible). Mass of calcite after cleaning was calculated from the Ca concentration of the analysed sample. Each sample consisted of between 21 and 52 tests from the 300– 355 μm size fraction. All samples were cleaned and analysed at the Godwin Laboratory, University of Cambridge and were cleaned according to the protocols used there. The Mg-cleaning method is based on Barker et al. (2003). The method for Cd-cleaning is based on that of Boyle (1981). There were some slight variations to the published methods, as follows. In both cleaning methods the “coarse silicate removal” step was carried out directly after the deionised water and methanol rinses. The reductive step was carried out before the oxidative step in the Cdcleaning method. In order to isolate the effect of reductive cleaning Mg-cleaned samples received an extra oxidative step in lieu of the reductive step. This ensured that the number of rinses and amount of sample manipulation was the same for the two sets of samples. After cleaning, samples were analysed, first by ICP-OES (Inductively Coupled Plasma-Optical Emission Spectrometry) to obtain Ca concentrations and also Fe/Ca. Samples were then diluted to a constant Ca concentration and analysed using the ICP-MS (Inductively Coupled PlasmaMass Spectrometry) method developed for B/Ca analysis (Yu et al., 2005). Long-term reproducibility for Mg/Ca is better than 1.5%. All the samples were cleaned and analysed over the same two week period. This manuscript deals with Ca, Mg, Sr, Fe, Mn and Al. Other element data (Li, B, Zn, Cd, Ba, U) for these samples exist and will be reported elsewhere (Yu, in preparation). At selected depths, an extra sample was cleaned (including leach step) for SEM (scanning electron microscope) examination to observe the physical effect of reductive cleaning on the tests. Samples were fixed to stubs and carbon coated before examination using a JEOL JSM 840 SEM in the Department of Geosciences, University of Bremen. In order to assess whether Mg/Ca values obtained from the two cleaning methods were more similar to each other from deep sites we compare regressions between Δ[CO23 −] and Mg/Ca. Regressions be2− tween parameters (Δ[CO2− 3 ] and, Mg/Ca; also Δ[CO3 ] and analytical yield) were estimated by ordinary least squares (OLS) bisector method (Isobe et al., 1990; Fiegeson and Babu 1992). This was selected as it does not assume that there is an independent and dependent variable, but that uncertainty is equally likely on both parameters. Uncertainty on Δ[CO2− 3 ] arises if the samples are subject to porewater, rather than bottom water. Uncertainty on Mg/Ca arises due to natural variability and changes in temperature or habitat depth as well as instrumental error of the Mg/Ca analysis. The effect of reductive cleaning on yield and on element to Ca ratios (Mg/Ca, Sr/Ca, Mn/Ca, Fe/Ca, Al/Ca) was evaluated by comparison of

pairs of samples where each pair consisted of one Mg-cleaned and one Cd-cleaned sample from the same depth. Literature suggests that Cdcleaning tends to result in lower values of Mg/Ca, Mn/Ca and Fe/Ca (eg Barker et al. (2003); Weldeab et al. (2006)) than Mg-cleaning, so the former was subtracted from the latter. Differences between the pairs were averaged for all depths in a sample. This average difference were considered statistically significant when a one sided t-test gave p b 0.1. When this was the case, the null hypothesis (reductive cleaning does not result in a lower value) was rejected. Samples from a core taken from deep (4157 m) in the Indian Ocean WIND28K (10° 09.2′ S, 51° 46.2′ E) (McCave, 2001) were used to further examine the relationship between analytical yield and dissolution. 3. Results 3.1. Observations from scanning electron microscopy We identify the effects of reductive cleaning by comparing SEM of Mg-cleaned and Cd-cleaned samples. Corrosion caused by reductive cleaning is superimposed on the effects of natural dissolution at the seafloor plus any additional dissolution caused by Mg-cleaning. Corrosion is shown as cracking and roughening of the texture and the development of porosity in the test calcite. In the following text ‘well-preserved’ tests come from site 1BC3 at 1616 m on the OJP. SEM showed that after Mg-cleaning, pores of G. ruber (not shown), G. sacculifer (Fig. 1a) and N. dutertrei (Fig. 2c, e) sometimes contained coccoliths or other sediment. Cd-cleaned tests tended to have empty pores (Figs. 1b, 2b and f) and coccoliths were seen only very occasionally (Fig. 3b). 3.1.1. G. ruber and G. sacculifer The outer surface of well-preserved G. ruber and G. sacculifer was corroded by reductive cleaning. Fig. 1(a–d) shows well-preserved G. sacculifer (from 1616 m). The outer test surface is more etched after Cd-cleaning (Fig. 1b) than after Mg-cleaning (Fig. 1a). Side view of the test wall shows that the inner test in particular is more dissolved after Cd-cleaning (Fig. 1d) than Mg-cleaning (Fig. 1c). Reductive cleaning caused further corrosion of G. ruber and G. sacculifer tests which had already undergone partial dissolution at the seafloor. Fig. 1f shows that the outer test surface of G. sacculifer from 3411 m has more cracks after Cd-cleaning than after Mg-cleaning (Fig. 1e). Test wall of G. ruber from 2301 m shows slight separation between inner layers of calcite after Mg-cleaning (1 g); the wall appears yet more dissolved and porous after reductive cleaning (1 h). 3.1.2. N. dutertrei Reductive cleaning caused slight etching of the outer calcite of wellpreserved tests. Fig. 2 shows that the outer surface of reductively cleaned tests (Fig. 2b) were more pitted, particularly around the pores, than Mg-cleaned ones (Fig. 2a). Enhanced corrosion was more evident in the inner calcite. Inner calcite of well-preserved N. dutertrei shows slight separation of layers after Mg-cleaning (Fig. 2c); corrosion of the inner calcite was more apparent after reductive cleaning (Fig. 2d). Inner calcite of Mg-cleaned tests shows damage mainly around the pores (Fig. 2e), whereas Cd-cleaned tests show more widespread etching of the inner calcite (Fig. 2f). The outer surface of poorly-preserved N. dutertrei from deep sites did not appear to be further corroded by reductive cleaning. Fig. 2g and h shows that the outer surface of tests from 34,011 m are etched and pitted after both cleaning methods. Inner calcite of N. dutertrei from deep sites was porous and fissured or, particularly after reductive cleaning, missing. Fig. 2i shows corroded inner calcite of Mg-cleaned test from 3411 m. Fig. 2j shows reductively cleaned test from the same site where the inner calcite layers are missing..

H.J.H. Johnstone et al. / Chemical Geology 420 (2016) 23–36

27

Fig. 1. SEM image of cleaned G. ruber and G. sacculifer tests from the OJP. Left hand side: tests cleaned using Mg-cleaning method; right hand side: tests cleaned using Cd-cleaning method. Values in () are Δ[CO2− 3 ] in μmol/kg. White scale bars are 10 μm long. (a–d) G. sacculifer from 1616 m (13.8), (a) and (b) outer test; (c) and (d) side view of test wall. (e–f) G. sacculifer from 3411 m (− 5.8), outer test. (g–h) G. ruber from 2301 m (4.9), side of test wall.

3.1.3. P. obliquiloculata Well-preserved tests showed very slight etching of the outer surface after reductive cleaning; calcite of the inner test was more affected. Fig. 3 shows that inner calcite of Cd-cleaned test (Fig. 3b) is more dissolved than that of Mg-cleaned test (Fig. 3a). Fig. 3d shows close-up of Fig. 3b, the very outer layer of calcite appears less corroded than the inside. Poorly preserved tests of P. obliquiloculata tended to split into layers. Fig. 3c shows tests from 2301 m; the outer veneer has separated from the rest of the test. Inner calcite of poorly preserved tests appeared friable and porous after both cleaning methods, but more so after reductive

Fig. 2. SEM images of cleaned N. dutertrei tests from the OJP. Left hand side: tests cleaned by Mg-cleaning method; right hand side: tests cleaned using Cd-cleaning method. Values in () are Δ[CO2− 3 ] in μmol/kg. White scale bars are 10 μm long. (a–f) N. dutertrei from 1616 m (13.8), (a) and (b) outer test; (c) and (d) side view of test wall; (e) and (f) inner test. (g–j) N. dutertrei from 3411 m (− 5.8), (g) and (h) outer test; (i) and (j) inner test.

28

H.J.H. Johnstone et al. / Chemical Geology 420 (2016) 23–36

due to reductive cleaning did not quite reach statistical significance for G. sacculifer (0.10 mmol/mol, or 3%, α = 0.10). In this species, there was a significant difference in Mg/Ca between cleaning methods in samples from above the calcite saturation horizon, where Mg/Ca was on average 0.21 mmol/mol, ~5%, less for Cd-cleaned rather than Mg-cleaned samples. There was no difference in the offset between methods in samples from above and below the calcite saturation horizon for any other species. There was no decrease in average Mg/Ca between the two cleaning methods for P. obliquiloculata. The highest values of Al/Ca, Fe/Ca and Mn/Ca were found in Mg-cleaned samples from above the calcite saturation horizon. Samples from deep sites at the Ontong Java Plateau tended to have lower levels of contamination particularly Fe/Ca and Mn/Ca. Mean Fe/Ca and Mn/Ca were lower after reductive cleaning for all four species (Fig. 5). The ICP-OES recorded negative values for Fe for most of the reductively cleaned samples, meaning values were below the level of the (dissolution acid used as the) blank. Response of the ICP-OES to Fe is highly non-linear at low values, and where negative Fe values were recorded a value of zero has been used for Fe/Ca (Table 2) in order to compare samples. Comparison of sample means showed that reductive cleaning lowered Fe/Ca by 93% for G. ruber; 94% for G. sacculifer; 72% for N. dutertrei. There was no significant decrease for P. obliquiloculata. A negative value was recorded for Mn for several of the reductively cleaned samples, indicating that concentration was below the value of the (dissolution acid used as the) blank on the ICPMS. In this case, recorded values are used for the calculations. Reductive cleaning lowered mean Mn/Ca by 83% for G. ruber; 97% for G. sacculifer; 81% for N. dutertrei and 53% for P. obliquiloculata. Mean Al/Ca was lower after reductive cleaning by 71% for G. ruber; 23% for G. sacculifer and 66% for N. dutertrei. Cleaning method made no difference to Al/Ca of P. obliquiloculata. 3.3. Relationship of yield with Δ[CO2− 3 ] and with Mg/Ca

Fig. 3. SEM images of cleaned P. obliquiloculata tests from the OJP. Left hand side: tests cleaned using Mg-cleaning method; right hand side: tests cleaned using Cd-cleaning method.

cleaning. Fig. 3e shows corrosion of inner calcite in a Mg-cleaned test from 2965 m; inner calcite appears even more porous after reductive cleaning (Fig. 3f). 3.2. Geochemical analysis 3.2.1. Sensitivity of Mg/Ca to Δ[CO2− 3 ] Regressions between Mg/Ca and Δ[CO2− 3 ] (Fig. 4) were similar for both cleaning methods for G. ruber, N. dutertrei and P. obliquiloculata. Mg/Ca of Cd-cleaned G. sacculifer appeared less sensitive to Δ[CO23 −] than Mg-cleaned samples (Fig. 4). 3.2.2. Effect of reductive cleaning on Mg/Ca, indicators of contamination (Al/Ca, Fe/Ca and Mn/Ca), pairwise comparisons Thirty-one pairs of samples were compared, where each pair consisted of one Mg-cleaned and one Cd-cleaned sample from the same depth. Sample pairs came from 6 depths for G. ruber, 7 depths for G. sacculifer, 9 depths for N. dutertrei and 9 depths for P. obliquiloculata (though for this species one sample (2.5BC37) was lost during cleaning). Four extra samples were also run which did not have a partner. The average of the differences between pairs (Table 3), showed that Cd-cleaning resulted in lower Mg/Ca than Mg-cleaning for two of the four foraminifera species. Mean difference in Mg/Ca was 0.17 mmol/mol, or ~10%, lower after Cd-cleaning than Mg-cleaning for N. dutertrei and 0.18 mmol/mol, ~ 4%, lower for G. ruber. The decrease in mean Mg/Ca

Analytical yield (Table 3, Fig. 6) was greater after Mg-cleaning than after Cd-cleaning for G. ruber (Mg-cleaning 32%; Cd-cleaning 16%), G. sacculifer (37%; 29%) and N. dutertrei (36%; 25%). Yield was not significantly different between methods for P. obliquiloculata (27%; 35%). Regressions estimated between analytical yield of core top samples and Δ[CO2− 3 ] all show a positive relationship. Correlation was not significant at the 10% level for G. ruber and P. obliquiloculata cleaned by the Mgmethod. The relationship may be mediated by sample mass. There is a strong correlation between Δ[CO23 −] and mass of the initial sample, and analytical yield is moderately correlated to sample mass. For the down core record, WIND28K, yield and dissolution index, XDX, of G. sacculifer show a weak but statistically significant correlation (r = −0.32, p = 0.020). Correlation was stronger when the dataset was restricted to values of yield N30% (r = −0.61, p = 0.002). Correlation between sample mass and dissolution index XDX was strong (r = − 0.78, p = b 0.001). There was some correlation between yield and total sample mass (r = 0.21, p = 0.130) and yield and average mass (r = 0.31, p = 0.030). 3.3.1. Mg/Ca and yield Correlation between yield and Mg/Ca for the OJP core-top samples was less than that between Δ[CO23 −] and Mg/Ca (Fig. 4). Correlation was significant, p b 0.050, for N. dutertrei and P. obliquiloculata (Table 3). 4. Discussion 4.1. Cause of decreased Mg/Ca after reductive cleaning: selective dissolution or more effective contaminant removal Previous studies using SEM to record dissolution features of planktonic forams describe how the outer surface of porous species (such as G. ruber, G. sacculifer and N. dutertrei) becomes pitted and dissolved by corrosive waters, while the smooth surface of P. obliquiloculata becomes

H.J.H. Johnstone et al. / Chemical Geology 420 (2016) 23–36

29

Fig. 4. Mg/Ca for four species of planktic foraminifera prepared by Mg-cleaning (blue triangles) and Cd-cleaning (red circles). Sensitivity of Mg/Ca to Δ[CO2− 3 ] (decrease in mmol/mol per μmol/kg) is shown for each species and cleaning method. Cd-cleaned data of Dekens et al. (2002) (grey empty circles) shown for comparison, note different size fractions. Shaded areas at the top of the graphs (above 15 μmol/kg) show theoretical Mg/Ca calculated from water temps at the OJP. On the G. ruber and G. sacculifer graph, the shaded areas represent seasonal range in SST of 28.9–29.3 °C (Locarnini et al., 2006). Thermocline temperatures at OJP are between 11.9 to 28.9 °C (Locarnini et al., 2006), blue shaded areas on the N. dutertrei and P. obliquiloculata graphs illustrate average thermocline temperature of 11.9 °C ± 0.5 °C. These temperatures were calculated in terms of Mg/Ca using published calibrations for Mg-cleaned (blue shading) (Anand et al., 2003) and Cd-cleaned (red shading) (Dekens et al., 2002) forams.

rougher. Microscopic cracks appear in the test wall and layers of calcite peel off, particularly around the pores (Bé et al., 1975; Dittert and Henrich, 2000; Johnstone et al., 2010). CT shows that inner calcite is preferentially dissolved (Johnstone et al., 2010; Iwasaki et al., 2015). Dissolution at the seafloor is known to decrease mass (e.g. Lohmann, 1995) and Mg/Ca (e.g. Brown and Elderfield, 1996; Regenberg et al., 2006) of foraminifera tests. Mass and (Mg-cleaned) Mg/Ca data have been previously published for the OJP; both quantities decrease linearly

with Δ[CO2− 3 ] (Johnstone et al., 2010; Johnstone et al., 2011) (Table 2 shows the subset of that data which is used in this present study). In the present study, comparison of pairs of Mg-cleaned and Cdcleaned samples from the same depth allows the effect of reductive cleaning on these samples with varying preservation states to be isolated. The enhanced physical corrosion of test calcite revealed by SEM, and the decrease in analytical yield, of Cd-cleaned compared to Mg-cleaned samples for three out of four species (Figs. 1, 2, 3, 6; Table 2) support

30

H.J.H. Johnstone et al. / Chemical Geology 420 (2016) 23–36

Table 3 2− 2− Parameters for regressions between Mg/Ca and Δ[CO2− 3 ] (Fig. 4) and analytical yield and Δ[CO3 ] (Fig. 6). Also correlation (r) between sample mass and Δ[CO3 ], and partial correlation between yield and Δ[CO2− 3 ], controlling for sample mass, for core-top samples from the OJP. Species Core-tops

Cleaning protocol

G. ruber G. sacculifer N. dutertrei P. obliquiloculata G. ruber G. sacculifer N. dutertrei P. obliquiloculata

Mg-cleaning Mg-cleaning Mg-cleaning Mg-cleaning Cd-cleaning Cd-cleaning Cd-cleaning Cd-cleaning

n

6 7 10 10 6 7 10 10

Regression between cleaning yield and Δ[CO2− 3 ] (Fig. 6)

Slope

Intercept

Slope

Intercept

r

p

r

p

r

p

24 ± 4 24 ± 6 30 ± 2 25 ± 1 20 ± 4 49 ± 5 35 ± 4 26 ± 2

−101 ± 20 −88 ± 22 −54 ± 4 −56 ± 3 −77 ± 18 −176 ± 19 −56 ± 7 −59 ± 4

0.89 ± 0.15 0.59 ± 0.15 0.82 ± 0.11 0.71 ± 0.08 0.53 ± 0.07 0.54 ± 0.15 1.29 ± 0.37 0.97 ± 0.18

−24 ± 6 −24 ± 8 −32 ± 5 −24 ± 5 −4 ± 2 −13 ± 5 −36 ± 9 −37 ± 7

0.99 0.86 0.84 0.93 0.92 0.66 0.83 0.97

0.000 0.013 0.002 0.000 0.009 0.110 0.003 0.000

0.51 0.64 0.41 0.58 0.67 0.26 0.58 0.80

0.301 0.125 0.238 0.079 0.149 0.571 0.079 0.006

−0.89 0.40 0.64 −0.16 0.91 0.80 0.26 −0.66

0.020 0.380 0.050 0.650 0.010 0.030 0.470 0.040

Gravity core WIND28K

G. sacculifer

Mg- cleaning

52 23

Partial correlation between yield and Δ[CO2− 3 ]

Regression between Mg/Ca and Δ[CO2− 3 ] (Fig. 4)

Whole sample set Subset where yield N30%

Correlation between sample mass and Δ[CO2− 3 ]

Correlation between sample mass and yield

Regression between yield and XDX

Correlation between sample mass and XDX

Correlation between sample mass and yield

Partial correlation between yield and Δ[CO2− 3 ]

r

p

r

p

r

p

r

p

−0.32 −0.61

0.020 0.002

−0.78

b0.001

0.21

0.130

0.26 0.73

0.063 b0.001

previous reports that reductive cleaning causes dissolution of foraminiferal calcite (Barker et al., 2003; Yu et al., 2007; Marr et al., 2013; Vetter et al., 2013; Sadekov et al., 2010). SEM shows that the inner calcite of N. dutertrei and P. obliquiloculata was more corroded by reductive cleaning than was the outer crust (Figs. 2b and d, 3b and d). Chemical mapping has shown that N. dutertrei and P. obliquiloculata species have a low Mg outer crust (Sadekov et al., 2005; Kunioka et al., 2006; Sadekov et al., 2010) and preferential dissolution of inner calcite has been explained as selective dissolution of Mg-rich parts (Lorens et al., 1977; Russell et al., 1994; Brown and Elderfield, 1996; Hastings et al., 1998; Rosenthal et al., 2000). However, this ‘selective dissolution’ did not necessarily lead to change in Mg/Ca. P. obliquiloculata tests were corroded by reductive cleaning (Fig. 3), yet there was no significant difference in Mg/Ca between cleaning methods in this species. It appears that tests can be partially dissolved by reductive cleaning without this significantly affecting Mg/Ca. Addressing the suggestion of Barker et al. (2003) outlined in the Introduction, there was no difference in the slopes of regressions between Δ[CO2− 3 ] and Mg/Ca of Mg-cleaned and Cd-cleaned samples for three out of the foraminiferal species analysed. The exception was G. sacculifer, where OLS bisector regressions trended toward smaller offset for poorly preserved samples. In this species, Mg/Ca of Cd-cleaned samples appeared less sensitive to the effect of dissolution, decreasing by 0.021 (± 0.002) mmol/mol per unit Δ[CO23 −] compared to 0.041 (±0.010) for Mg-cleaned samples. In our study, the average (for all depths) decrease in Mg/Ca due to reductive cleaning, was less than the general correction factor of 15% suggested by Rosenthal et al. (2004), for all four species examined. The average offset for G. ruber white in this study of 4% (Fig. 4, Fig. 5) is less than the 15% found by Barker et al. (2003) for well-preserved G. ruber white from the Arabian Sea. We cannot completely exclude the possibility of a preservation effect on our ruber samples. It may be that our sample set does not encompass truly well-preserved samples, G. ruber from even the shallowest sites on the Ontong Java Plateau appears slightly dissolved [Johnstone et al., 2010] and Mg/Ca of the shallowest samples does not represent annual average SST at the OJP (Fig. 4). Therefore, we cannot completely exclude the possibility that there is a preservation effect in this species, although, for the range of dissolution covered by our sample set, there was no trend toward a smaller offset with decreased calcite saturation. Alternatively, there are study specific differences in how much reductive cleaning decreases

Mg/Ca. Xu et al. (2008) also found a smaller offset (6.5%) for G. ruber than did Rosenthal et al. (2004) and Barker et al. (2003). For G. sacculifer, there was no significant difference in average Mg/Ca of the Mg-cleaned and Cd-cleaned sample set. This is in agreement with Barker et al. (2003) who also found no difference in this species, but differs from Rosenthal et al. (2004) who found 8%. Considering samples above the calcite saturation horizon only, Mg/Ca was 5% lower after Cd-cleaning. The effect on Mg/Ca of dissolution during cleaning is not predicted by dissolution at the seafloor. N. dutertrei and P. obliquiloculata from the Ontong Java Plateau show a similar sensitivity to the effect of natural dissolution on Mg/Ca (Fig. 4). However, on average, Mg/Ca of N. dutertrei was ~ 10% lower after reductive cleaning, suggesting that this species is very sensitive to cleaning protocol, while that of P. obliquiloculata was unaffected by cleaning method (Fig. 4). Again this differs from Rosenthal et al. (2004) who found an 8% decrease with Cd-cleaning in P. obliquiloculata. Cd-cleaned Mg/Ca data of Dekens et al. (2002) is shown on Fig. 4 for comparison. Regressions between Δ[CO2− 3 ] and Mg/Ca of that data are statistically indistinguishable from those of our Cd-cleaned data. Absolute values for G. ruber and G. sacculifer are also similar, but Mg/Ca of N. dutertrei is offset. A possible explanation for this may be that the Dekens et al. (2002) data comes from a different (smaller) size fraction. Friedrich et al. (2012) show that there is an inverse relationship between test size and Mg/Ca for some species. It may be that N. dutertrei is particularly sensitive to this effect. Efficiency of contaminant removal may contribute to the difference in Mg/Ca between cleaning methods. Although these OJP core top samples were cleaned adequately by Mg-cleaning (Fe/Ca and Al/Ca levels were never N0.1 mmol/mol as suggested by Barker et al. (2003) to indicate contamination), SEM of our samples shows that Cd-cleaning is a more effective method. Mg-cleaning can leave coccolith plates and other detritus in test pores (Fig. 1a, 2c), while reductively cleaned samples had empty pores. Coccoliths would not themselves contribute significant Mg as they tend to have low Mg/Ca (Stoll et al., 2001). Decreased Al/Ca after reductive cleaning in G. ruber, G. sacculifer and N. dutertrei in this study (Fig. 5) also confirms the work of Boyle (1981) that Cd-cleaning removes sediment from tests more effectively than Mg-cleaning. It may be that the slight calcite dissolution associated with reductive cleaning is actually advantageous, and releases authogenic calcites and particles adsorbed to the test. Metal oxide phases are more likely to occur in down core samples, but even in

H.J.H. Johnstone et al. / Chemical Geology 420 (2016) 23–36

31

Fig. 5. Comparison of element/Ca obtained by Mg-cleaning and Cd-cleaning for four species of planktic foraminifera. Solid lines are 1:1 lines; dashed line best fit to data forced through the origin. Where analyte was below the calibration for Fe, a value of zero has been plotted for Fe/Ca. On average (all samples, all species) Mg/Ca is 4% lower where Cd-cleaning rather than Mg cleaning was employed. Fe/Ca and Mn/Ca are lower after reductive cleaning for all four species. Al/Ca is lower after reductive cleaning for all species except P. obliquiloculata.

these Pacific core-top samples Mn/Ca was lower after Cd-cleaning (Fig. 5). These oxides can, in some circumstances, contribute contaminant Mg (Pena et al., 2005; Weldeab et al., 2006). Potentially, they could also act as a cement, the solution of which would release contaminant particles. The tendency of tests to collect sediment or other contaminants may depend on morphology. Barker et al. (2003) demonstrated that G. bulloides, with its porous texture and open form, was more prone to clay contamination than other species. Previous studies have noted that pores are sites of contaminants (Pena et al., 2008; Vetter et al., 2013). In this present study, the smooth surface of P. obliquiloculata (Fig. 3d) was never seen to be contaminated with sediment, unlike the porous tests of G. ruber, G. sacculifer (Fig. 1) and N. dutertrei (Fig. 2). In contrast to the three species above, Al/Ca (and Mg/Ca) of P. obliquiloculata was similar for both cleaning methods (Fig. 5) suggesting that there was little clay to be removed. We do not suggest that clay

removal accounts for all of the differences in Mg/Ca between cleaning methods as what is removed does not fit the element ratio of typical marine clays. For example illite is roughly 2% Mg, 2% Fe, 10% Al, i.e. contains ~5 times more Al than Mg, whereas reductive cleaning removed ~6–7 times more Mg than Al or Fe (Table 3). The study specific, rather than species specific, differences in Mg/Ca between cleaning methods support the supposition of Rosenthal et al. (2004) that minor differences in cleaning protocol affect Mg/Ca values. In this study the “silicate removal step” was carried out for all samples, and in some laboratory protocols this step is omitted. Additionally the “oxidative step” was carried out twice in the Mg-cleaning method. We speculate that the intensity of sample crushing may also affect results. Recent studies using LA show that dissolution thins the walls of tests, but that high Mg bands persist in partially dissolved G. ruber (Tachikawa et al., 2008), G. sacculifer (Sadekov et al., 2010), O. universa (Vetter et al., 2013; Sadekov et al., 2010), and G. bulloides

32

H.J.H. Johnstone et al. / Chemical Geology 420 (2016) 23–36

Fig. 6. Correlation (r) between analytical yield and deep water calcite saturation, Δ[CO2− 3 ]. For G. ruber, G. sacculifer and N. dutertrei mean analytical yield is lower after reductive cleaning. In P. obliquiloculata yield is similar for both cleaning methods. Slopes of regressions are given in Table 3.

(Marr et al., 2013). Finer crushing potentially exposes more Mg-rich areas to corrosive reagents. Sensitivity to reagents may also be influenced by intrinsic characteristics of the test, such as arrangement of Mg or crystallinity, which vary between samples. Likelihood of contamination and the amount of Mg in contaminant phases also varies with locality. A universal correction factor therefore may not be appropriate. Where high accuracy is required, comparisons should be made for the particular protocols and samples involved. 4.2. Cleaning yield as a potential indicator of sample dissolution and bias of Mg/Ca derived temperatures Maximum yield from core-top samples was ~ 60%, so a significant part of the original sample was never analysed, even in well-

preserved tests cleaned by the gentlest (Mg-cleaning) method. Presumably part of the not-analysed fraction is clay, silicates and other contaminants. SEM showed that samples from below the calcite saturation horizon contained less detritus than those from above it. They also had lower Al/Ca (for Mg-cleaned samples: G. ruber was 58% lower below CSH compared to above CSH, G. sacculifer 20% lower, N. dutertrei 48% lower, P. obliquiloculata 2% lower). Since more material in total was lost from samples from below the CSH, most of the loss of material in poorly preserved samples therefore must be from the tests themselves. Samples from deep sites were noticeably more friable than well-preserved tests. They were easy to crush and formed small powdery fragments which remained on the surface of the cleaning solution and could potentially be discarded with the solution.

H.J.H. Johnstone et al. / Chemical Geology 420 (2016) 23–36

4.2.1. Yield as dissolution indicator Correlation between Δ[CO2− 3 ] and analytical yield (Table 3) confirms anecdotal evidence that yield tends to be lower for poorly preserved samples (Fig. 6, Table 3). In order to further test the relationship between yield and test preservation, we compared analytical yield of (Mg-cleaned) G. sacculifer to a record of preservation for a down core record. Sediment core WIND28K (McCave, 2001) spans the past 150 ka. It was retrieved from 4175 m water depth in the Indian Ocean from a site currently bathed in corrosive Circumpolar Deep Water. Δ[CO2− 3 ] proxy, XDX, is based on the appearance of tests in CT scans. The XDX record shows that tests are poorly preserved throughout much of the core (Johnstone et al., 2014) (Fig. 7). Good preservation (low XDX values) in WIND28K occurs during the deglaciations — generally times of good calcite preservation in Indian and Pacific Oceans (e.g. Berger, 1977) – and early in Marine Isotope Stage 3 – also a period of enhanced calcite preservation in the deep Indian Ocean (Anderson et al., 2008).

33

Good preservation over the deglaciation is also shown in the fragmentation index (MFI) record of Mekik et al. (2012) which confirms that calcite is well preserved between 9 and 21 ka. There is a weak but significant correlation between yield and preservation state as indicated by XDX, (r = −0.32, p = 0.020). Considering only a subset of the data, where yield is N 30%, gives a stronger correlation (r = −0.61, p = 0.002). This suggests that low yield can occur for any sample, but high yield tends to be associated with better preserved samples. Running correlation (over 7 datapoints) in Fig. 7 shows that the most significant correlation between the two parameters occurs during the steady changes in preservation of the deglaciations. It is possible that human factors may distort the relationship between yield and dissolution if poorly preserved or small samples were treated more carefully than samples where tests appear more robust. The WIND28K record allows examination of whether the sample mass is a modifying factor in the relationship between Δ[CO23 −] and

Fig. 7. Analytical yield for Mg-cleaned G. sacculifer from WIND28K, a core from deep (4157 m) in the Indian Ocean. Black line is δ18O of C. wuellerstorfi (McCave et al., 2005). Red line is dissolution index XDX (Johnstone et al., 2014), indicating preservation state of G. sacculifer tests. Good preservation and low XDX values, occur during the glacial terminations (marked TI and TII) and between 150 and 270 cm. Brown line is analytical yield of Mg-cleaned G. sacculifer (thick line is 3 point running average). Horizontal lines are running (7 values) correlation coefficient between XDX and % yield (dark green) and p value (light green). Yellow bars highlight areas where there is significant correlation, r b −5 and p b 0.01. Lower panel shows regression between XDX and % yield, where yield is N30%.

34

H.J.H. Johnstone et al. / Chemical Geology 420 (2016) 23–36

analytical yield. In OJP core-top samples there was strong correlation between sample mass and yield at least in some species (e.g. P. obliquiloculata r = 0.80, p = 0.006, Table 3). Sample mass is also strongly correlated to Δ[CO23 −] in core-tops (Table 3) as test walls thin, and tests become less frequent in the sediment, due to calcite dissolution. However, in the down core record, WIND28K, correlation between sample mass and yield for (Mg-cleaned) G. sacculifer was less strong (r = 0.21, p = 0.130) than correlation between sample mass and yield for (Mg-cleaned) G. sacculifer from the core-top samples (r = 0.64, p = 0.125). In a record covering many thousands of years, sample mass probably changes for reasons other than Δ[CO2− 3 ], and analytical yield thus offers an independent way to assess dissolution. 4.2.2. Does yield indicate altered Mg/Ca? The effect of dissolution on Mg/Ca derived temperatures is a major problem for paleoceanographic reconstructions — for instance, dissolution reduces temperatures calculated from N. dutertrei by ~9 °C from the shallowest to deepest samples from the Ontong Java Plateau (Johnstone et al., 2011). Our core-top samples suggest that for Mg-cleaned samples, Mg/Ca is not reliable where yield b30%. Mg/Ca of these samples was biased by at least 10%, which is equivalent to ~1 °C (Fig. 8). For Cd-cleaned samples the picture is less clear cut. G. ruber and G. sacculifer yield can be very low (b10%) while Mg/Ca is still representative. For the more robust species, N. dutertrei and P. obliquiloculata, discarding Mg/Ca where yield was b 30% would not get rid of all biased values and would reject one good value. Despite the caveats discussed above, and the probability that the value of yield below which data is unreliable is likely to vary between workers, we agree with Tachikawa et al. (2008) that it is worth monitoring yield as a first indicator of dissolution bias on Mg/Ca derived temperatures. Although there is a great deal of variability between individual data points, a succession of low yields may warn of a dissolved section of core. 5. Conclusions (1) There was no difference in the slope of regression between Δ[CO23 −] and Mg/Ca of Mg-cleaned and Cd-cleaned samples for G. ruber, N. dutertrei or P. obliquiloculata. In these species there was no preservation effect in the offset in Mg/Ca between the two cleaning methods. For G. sacculifer, Mg/Ca of Cd-cleaned samples appeared less sensitive to the effect of dissolution. Although Mg-cleaning is an adequate cleaning method, at least for these Pacific samples, which are not heavily contaminated, Cdcleaning removes contaminants more effectively, especially for porous species. SEM showed that there was sometimes detritus in pores of G. ruber, G. sacculifer and N. dutertrei after Mg-cleaning while Cdcleaned tests had clean and empty pores. Elements indicative of contamination, Al/Ca, Fe/Ca and Mn/Ca, were lower after Cd-cleaning in these three species. Susceptibility to contamination is sensitive to morphology. P. obliquiloculata, which has a very smooth outer surface, did not have coccolith plates or sediment adhering to tests. Al/Ca was not decreased by reductive cleaning in this species, suggesting that there was little contamination to be removed. Reductive cleaning causes slight dissolution of foraminifera tests. Lower analytical yield in G. ruber, G. sacculifer and N. dutertrei species after Cd-cleaning confirms this as the more aggressive method. Even tests which had already undergone dissolution at the sea floor showed additional corrosion due to reductive cleaning, and that this corrosion was focused on inner, Mg-rich areas of the test. This slight dissolution did not necessarily decrease Mg/Ca. The inner test of P. obliquiloculata was corroded by Cd-cleaning, but Mg/Ca was not lower. Reductive cleaning did decrease Mg/Ca in some species. Mg/Ca was on average 4% lower for G. ruber (average of all depths), 5% lower for G. sacculifer (if only samples above the calcite saturation horizon are considered), and 10% lower for N. dutertrei (all depths) after reductive cleaning.

Fig. 8. Change in Mg/Ca due to dissolution (ΔMg/Ca) versus analytical yield for core top samples from the Ontong Java Plateau. Values from the shallowest sample (1BC3, 1616 m water depth) in our depth transect from the Ontong Java Plateau were used as the 100% value. The red shaded area shows where Mg/Ca is biased by more than 10% (equivalent to ~1 °C). Where analytical yield is below 30% (red horizontal line) the majority of Mg/Ca values are biased by dissolution.

These offsets are generally lower than those of previous studies. If the decrease in Mg/Ca due to reductive cleaning is controlled by extrinsic factors, such as amount or type of clay contamination or intrinsic factors of the test calcite, which can change through time, or to slight differences in cleaning protocol a universal, or even species specific, correction factor between methods would not apply. (2) Weak, but significant, correlation between analytical yield and Δ[CO2− 3 ] in core-top samples suggests that it may be worth monitoring this property as a first indicator of test preservation and dissolution bias on Mg/Ca. Values are no doubt operator dependent, but in this study, Mg/Ca values of Mg-cleaned samples were unreliable below 30% recovery. Analytical yield of Cd-cleaned samples, particularly of fragile species, can be low (b10%) but still provide reliable Mg/Ca.

H.J.H. Johnstone et al. / Chemical Geology 420 (2016) 23–36

White scale bars are 10 μm long. Black scale bar (c) is 100 μm long. (a), (b), (d) P. obliquiloculata from 1616 m (13.8), (a) (b) side view of test wall; (d) close-up of the area of (b) outlined in white. (c) Crushed cleaned sample from 2301 m (4.9). (e) and (f) Side of test wall from 2965 m (−2.3). The equations of the latter contain a Δ[CO2− 3 ] term, to account for dissolution effects, for which a value of 15 μmol/kg has been used. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Acknowledgements This study was funded through DFG Research Center / Cluster of Excellence “The Ocean in the Earth System” (FZT15/EXC309). Many thanks to Jimin Yu for running samples on the ICP-MS and for the helpful discussion. Thanks also to Mervyn Greaves for the help in the laboratory, and to Linda Booth, Stijn de Schepper and Henning Kuhnert for their useful comments. WL acknowledges the support of the Mathematics Applications Consortium for Science and Industry funded by the Science Foundation Ireland mathematics initiative grant 06/MI/005. References Anand, P., Elderfield, H., Conte, M.H., 2003. Calibration of Mg/Ca thermometry in planktonic foraminifera from a sediment trap time series. Paleoceanography 18 (2), 1050. http://dx.doi.org/10.1029/2002PA000846. Anderson, R.F., Fleisher, M.Q., Laoc, Y., Winckler, G., 2008. Modern CaCO3 preservation in equatorial Pacific sediments in the context of late-Pleistocene glacial cycles. Mar. Chem. 111, 30–46. Barker, S., Greaves, M., Elderfield, H., 2003. A study of cleaning procedures used for foraminiferal Mg/Ca paleothermometry. Geochem. Geophys. Geosyst. 4 (9), 8407. http:// dx.doi.org/10.1029/2003GC000559. Bé, A.W.H., Morse, J.W., Harrison, S.M., 1975. Progressive dissolution and ultrastructural breakdown of planktonic foraminifera, dissolution of deep-sea carbonates, special publication no. 13. Cushman Foundation for Foraminiferal Research. Berger, W.H., 1977. Deep-sea carbonate and the late deglaciation preservation spike in pteropods and foraminifera. Nature 269, 301–304. Bian, N., Martin, P.A., 2010. Investigating the fidelity of Mg/Ca and other elemental data from reductively cleaned planktonic foraminifera. Paleoceanography 25, PA2215. http://dx.doi.org/10.1029/2009PA001796. Boyle, E., 1981. Cadmium, zinc, copper and barium in foraminifera tests. Geochim. Cosmochim. Acta 47, 1815–1819. Brown, S., Elderfield, H., 1996. Variations in Mg/Ca and Sr/Ca ratios of planktonic foraminifera caused by postdepositional dissolution — evidence of shallow Mg-dependent dissolution. Paleoceanography 11 (5), 543–551. Dekens, P.S., Lea, D.W., Pak, D.K., Spero, H.J., 2002. Core top calibration of Mg/Ca in tropical foraminifera: refining paleotemperature estimation. Geochemisty Geophysics Geosystems 3 (4), 1022. http://dx.doi.org/10.1029/2001GC000200. Dittert, N.R., Henrich, R., 2000. Carbonate breakdown in the South Atlantic: evidence from ultrastructure breakdown in G. bulloides. Deep-Sea Res. 47, 603–620. Elderfield, H., Ganssen, G., 2000. Reconstruction of temperature and δ18O of surface ocean waters using Mg/Ca of planktonic foraminiferal calcite. Nature 405, 442–445. Elderfield, H., Yu, J., Anand, P., Kiefer, T., Nyland, B., 2006. Calibrations for benthic foraminiferal Mg/Ca paleothermometry and the carbonate ion hypothesis. Earth Planet. Sci. Lett. 250 (3–4), 633–649. Feigelson, E.D., Babu, G.J., 1992. Linear regression in astronomy. II: The Astrophysical Journal 397, 55–67. Friedrich, O., Schiebel, R., Wilson, P.A., Weldeab, S., Beer, C.J., Cooper, M.J., Fiebig, J., 2012. Influence of test size, water depth, and ecology on Mg/Ca, Sr/Ca, δ18O and δ13C in nine modern species of planktic foraminifers. Earth Planet. Sci. Lett. 319–320, 133–145. Hastings, D.W., Russell, A.D., Emerson, S.R., 1998. Foraminiferal magnesium in Globeriginoides sacculifer as a paleotemperature proxy. Paleoceanography 13, 161–169. Hemleben, C., Spindler, M., Anderson, O.R., 1989. Modern Planktonic Foraminifera. Springer, New York (363 pp.). Isobe, T., Feigelson, E.D., Akritas, M.G., Babu, G.J., 1990. Linear regression in astronomy. I: The Astrophysical Journal 364, 104–113. Iwasaki, S., Kimoto, K., Sasaki, O., Kano, H., Honda, M.C., Okazaki, Y., 2015. Observation of the dissolution process of Globigerina bulloides tests (planktic foraminifera) by X-ray microcomputed tomography. Paleoceanography 30, 317–331. http://dx.doi.org/10. 1002/2014PA002639. Johnstone, H.J.H., Schulz, M., Yu, J., Elderfield, H., 2010. Inside story: an X-ray microtomography method for assessing dissolution in foraminifera tests. Marine Micropaleontogy 77, 58–70. Johnstone, H.J.H., Schulz, M., Yu, J., Elderfield, H., 2011. Calibrating computed tomography based dissolution index XDX to dissolution bias of Mg/Ca in planktic foraminifera. Paleoceanography 26, PA1215. http://dx.doi.org/10.1029/2009PA001902. Johnstone, H.J.H., Kiefer, T., Elderfield, H., Schulz, M., 2014. Calcite saturation, dissolutioncorrected foraminiferal test mass and dissolution-corrected Mg/Ca reconstructed

35

using XDX: a 150 ka record from the western Indian Ocean. Geochemistry, Geophysics, Geosystems 15 (3), 781–797. Kunioka, D., Shirai, K., Takahata, N., Sano, Y., Toyofuku, T., Ujiie, Y., 2006. Microdistribution of Mg/Ca, Sr/Ca, and Ba/Ca ratios in Pulleniatina obliquiloculata test by using a NanoSIMS: implication for the vital effect mechanism. Geochem.Geophys. Geosyst 7, Q12P20. http://dx.doi.org/10.1029/2006GC001280. Lea, D.W., Mashiotta, T.A., Spero, H.J., 1999. Controls on magnesium and strontium uptake in planktonic foraminifera determined by live culturing. Geochem. Cosmochim. Acta. 63, 2369–2379. Lea, D., 2004. The 100 000-yr cycle in tropical SST, greenhouse forcing, and climate sensitivity. J. Clim. 17, 2170. Locarnini, R.A., Mishonov, A.V., Antonov, J.I., Boyer, T.P., Garcia, H.E., 2006. World ocean atlas 2005. In: Levitus, S. (Ed.), Temperature vol. 1. NOAA Atlas NESDIS 61, U.S. Government Printing Office, Washington, D.C. (182 pp). Lohmann, G.P., 1995. A model for variation in the chemistry of planktonic foraminifera due to secondary calcification and selective dissolution. Paleoceanography 10, 445–457. Lorens, R.B., Williams, D.F., Bender, M.L., 1977. The early nonstructural chemical diagenesis of foraminiferal calcite. J. Sediment. Res. 47. http://dx.doi.org/10.1306/212F73C92B24-11D7-8648000102C1865D. Martin, P.A., Lea, D.W., 2002. A simple evaluation of cleaning procedures on fossil benthic foraminiferal Mg/Ca. Geochem. Geophys. Geosyst. 3 (10), 8401. http://dx.doi.org/10. 1029/2001GC000280. Marr, J.P., Bostock, H.C., Carter, L., Bolton, A., Smith, E., 2013. Differential effects of cleaning procedures on the trace element chemistry of planktonic foraminifera. Chem. Geol. 351, 310–323. McCave, I.N., 2001. R.R.S Charles Darwin cruise 129, report. Department of Earth Science. University of Cambridge, UK. McCave, I.N., Kiefer, T., Thornalley, D.R., Elderfield, H., 2005. Deep flow in the MadagascarMarscarene Basin over the last 150,000 years. Phil. Trans. R. Soc. 363, 81–99. McCorkle, D.C., Keigwin, L.D., 1994. Depth profiles of δ13C in bottom water and core top C. wuellerstorfi on the Ontong Java Plateau and Emperor Seamounts. Paleoceanography 9, 197–208. Mashiotta, T.A., Lea, D.W., Spero, H.J., 1999. Glacial–interglacial changes in subantarctic sea surface temperature and δ18O-water using foraminiferal Mg. Earth Planet. Sci. Lett. 170, 417–432. Mekik, A.F., Anderson, R., François, R., Loubere, P., Richaud, M., 2012. The mystery of the missing carbonate deglacial preservation maximum. Quat. Sci. Rev. 39, 60–72. http://dx.doi.org/10.1016/j.quascirev.2012.01.024. Nürnberg, D., 1995. Magnesium in tests of Neogloboquadrina pachyderma sinistral from high northern and southern latitudes. J. Foraminiferal Res. 25, 350–368. Nürnberg, D., Bijma, J., Hemleben, C., 1996. Assessing the reliability of Mg in foraminiferal calcite as a proxy for water mass temperatures. Geochim. Cosmochim. Acta 60, 803–814. Pena, L.D., Calvo, E., Cacho, I., Eggins, S., Pelejero, C., 2005. Identification and removal of Mn–Mg-rich contaminant phases on foraminiferal tests: implications for Mg/Ca past temperature reconstructions. Geochem. Geophys. Goesyst. 6 (9), Q09P02. http:// dx.doi.org/10.1029/2005GC000930 (ISSN: 1525-2027). Pena, L.D., Cacho, I., Calvo, E., Pelejero, C., Eggins, S., Sadekov, A., 2008. Characterization of contaminant phases in foraminifera carbonates by electron microprobe mapping. Geochem. Geophys. Goesyst 9 (7), Q07012. http://dx.doi.org/10.1029/ 2008GC002018 (ISSN: 1525-2027). Regenberg, M., Nürnberg, D., Steph, S., Groeneveld, J., Garbe-Schönberg, D., Tiedemann, R., Dullo, W.-C., 2006. Assessing the effect of dissolution on planktonic foraminiferal Mg/ Ca ratios: evidence from Caribbean core tops. Geochem. Geophys. Geosyst. (7), Q07P15 http://dx.doi.org/10.1029/2005GC001019. Regenberg, M., Regenberg, A., Garbe-Schönberg, D., Lea, D.W., 2014. Global dissolution effects on planktonic foraminiferal Mg/Ca ratios controlled by the calcite-saturation state of bottom waters. Paleoceanography 29, 127–142. http://dx.doi.org/10.1002/ 2013PA002492. Rosenthal, Y., Boyle, E.A., Slowey, N., 1997. Environmental controls on the incorporation of Mg, Sr, F and Cd into benthic foraminifera shells from Little Bahama Bank: prospects for thermocline paleoceanography. Geochim. Cosmochim. Acta 61, 3633–3643. Rosenthal, Y., Lohmann, G.P., Lohmann, K.C., Sherrell, R.M., 2000. Incorporation and preservation of Mg in Globigerinoides sacculifer: implications for reconstructing the temperature and 18O/16O of seawater. Paleoceanography 15 (1), 135. Rosenthal, Y., et al., 2004. Interlaboratory comparison study of Mg/Ca and Sr/Ca measurements in planktonic foraminifera for paleoceanographic research. Geochem. Geophys. Geosyst. 5, Q04D09. http://dx.doi.org/10.1029/2003GC000650. Russell, A.D., Emerson, S., Nelson, B.K., Erez, J., Lea, D.W., 1994. Uranium in foraminiferal calcite as a recorder of seawater uranium concentrations. Geochim. Cosmochim. Acta 58 (2), 671–681. Sadekov, A.Y., Eggins, S.M., De Deckker, P., 2005. Characterization of Mg/Ca distributions in planktonic foraminifera species by electron microprobe mapping. Geochem. Geophys., Geosyst 6, Q12P06. http://dx.doi.org/10.1029/2005GC000973. Sadekov, A.Y., Eggins, S.M., Klinkhammer, G.P., Rosenthal, Y., 2010. Effects of seafloor and laboratory dissolution on the Mg/Ca composition of Globigerinoides sacculifer and Orbulina universa tests — a laser ablation ICPMS microanalysis perspective. Earth Planet. Sci. Lett. 292, 312–324. Stoll, H.M., Ruiz Encinar, J., Ignacio Garcia Alonso, J., Rosenthal, Y., Probert, I., Klaas, C., 2001. A first look at paleotemperature prospects from Mg in coccolith carbonate: cleaning techniques and culture measurements. Geochem. Geophys. Geosyst. 2, 1047. http://dx.doi.org/10.1029/2000GC000144. Tachikawa, T., Sepulcre, S., Toyofuku, T., Bard, E., 2008. Assessing influence of diagenetic carbonate dissolution on planktonic foraminiferal Mg/Ca in the southeastern Arabian Sea over the past 450 ka: comparison between. Globigerinoides ruber and Globigerinoides

36

H.J.H. Johnstone et al. / Chemical Geology 420 (2016) 23–36

sacculifer, Geophys., Geosyst. 9 (4), Q04037. http://dx.doi.org/10.1029/2007GC001904 (ISSN: 1525-2027). Vetter, L., Spero, H.J., Russell, A.D., Fehrenbacher, J.S., 2013. LA-ICP-MS depth profiling perspective on cleaning protocols for elemental analyses in planktic foraminifers. Geochem. Geophys. Geosyst. 14 (8). http://dx.doi.org/10.1002/ggge.20163 (ISSN: 1525-2027). Weldeab, S., Schneider, R.R., Kölling, M., 2006. Comparison of foraminiferal cleaning procedures for Mg/Ca paleothermometry on core material deposited under varying terrigenous-input and bottom water conditions. Geochem. Geophys. Geosyst. (7), Q04P12 http://dx.doi.org/10.1029/2005GC000990. Xu, J., Holbourn, A., Kuhnt, W., Jian, Z., Kawamura, H., 2008. Changes in the thermocline structure of the Indonesian outflow during terminations I and II. Earth Planet. Sci. Lett. 273, 152–162.

Yu, J.M., Day, J., Greaves, M., Elderfield, H., 2005. Determination of multiple element/calcium ratios in foraminiferal calcite by quadrupole ICP-MS. Geochem. Geophys. Geosyst. 6, Q08P01. http://dx.doi.org/10.1029/2005GC000964. Yu, J.M., Elderfield, H., Greaves, M., Day, J., 2007. Preferential dissolution of benthic foraminiferal calcite during laboratory reductive cleaning. Geochem. Geophys. Geosyst. 8, Q06016. http://dx.doi.org/10.1029/2006GC001571. Yu, J.M., 2015. Dissolution Effects in Planktic Foraminifera From the Ontong Java Plateau (in preparation).