Detailed Abundances of Planet-hosting Open Clusters. The Praesepe (Beehive) Cluster^*

Stars in Praesepe are of special interest due to the discovery of planets around a number of its members (Quinn et al. 2012; Cai et al. 2019). Praesepe is home to ∼1000 stars, is relatively close by at 182 pc (Cantat-Gaudin et al. 2018), and has an age of ∼600 Myr (Delorme et al. 2011), making it a strong candidate for high-resolution spectroscopy of main-sequence Sun-like stars. It has the highest metallicity ([Fe/H] = +0.21 ± 0.01 dex) of any nearby open cluster according to D'Orazi et al. (2020), which can increase the likelihood of giant planet formation (Johnson et al. 2010) if the metallicity correlation applies to stars in open clusters.

The six stars in our sample consist of a planet host, Pr0201, and five non-hosts: Pr0133, Pr0081, Pr0076, Pr0051, and Pr0208. The data used in this study are a combination of spectra acquired by our group in 2013 and high-quality Keck Observatory Archive (KOA) spectra taken at an earlier time, all of which are publicly available. While other planet hosts exist within Praesepe, we only acquired data for Pr0201. The derived temperatures (see Sections 2.3 and 2.4) for our stars cover ∼500 K with spectral classes between G5 and F8 shown in Table 1. Half of our stars have temperatures within 100 K of the Sun and the other half are hotter with temperatures ∼6100 K. Five out of the six stars have similar estimates of surface gravity, ranging from 4.34 to 4.44 dex with errors of about 0.10 dex. Pr0133 is an exception, having a much lower surface gravity of 4.18 dex. The metallicity of our stars ranges between 0.16 and 0.26 with an average of +0.21 ± 0.02 dex. In Figure 1, we show fits to the broadband spectral energy distribution (using the methodology of Stassun & Torres 2016) for two of our stars, which are also consistent with our spectral analysis.

Zoom In Zoom Out Reset image size

Table 1. Stellar Parameters ^a

	Pr0201	Pr0133	Pr0208	Pr0081	Pr0051	Pr0076 (2003)	Pr0076 (2013)
Teff (K)	6168 ± 35	6067 ± 60	5869 ± 46	5731 ± 42	6017 ± 27	5789 ± 72	5748 ± 24
$\mathrm{log}g$ (cgs)	4.34 ± 0.10	4.18 ± 0.12	4.37 ± 0.13	4.44 ± 0.11	4.40 ± 0.07	4.48 ± 0.12	4.44 ± 0.07
[Fe/H]	0.23 ± 0.05	0.19 ± 0.06	0.26 ± 0.07	0.18 ± 0.06	0.16 ± 0.03	0.25 ± 0.06	0.22 ± 0.05
ξ (km s−1)	1.52 ± 0.06	1.74 ± 0.11	1.53 ± 0.07	1.41 ± 0.06	1.54 ± 0.05	1.33 ± 0.04	1.35 ± 0.03

Note.

^a Adopted solar parameters: T_eff = 5777 K, $\mathrm{log}g=4.44$, and ξ = 1.38 km s⁻¹.

Download table as:  ASCIITypeset image

Three of the six stars (Pr0201, Pr0051, and Pr0076) were observed on UT 2013 December 9 with the HIRES echelle spectrograph (Vogt et al. 1994) in the R = λ/Δλ = 72,000 mode on the 10 m Keck I telescope. We used the kv418 filter combined with the B2 slit setting (0574 × 7'') and 2 × 1 binning; the spectra cover a wavelength range of 4600–9000 Å. One exposure was taken for each star, with an integration time of 1200 s for Pr0201 and 2100 s for Pr0051 and Pr0076 individually. The signal-to-noise ratio (S/N) in the continuum near 6700 Å is ∼300 for Pr0201 and Pr0051, and ∼250 for Pr0076.

For the remaining three stars, Pr0133, Pr0208, and Pr0081, we obtained raw data files from the KOA. These spectra, which we will collectively refer to as archive spectra, were taken in UT 2003 January and February belonging to program ID H39aH and H47aH; the data are fully described in Boesgaard et al. (2013). We also obtained an archive spectrum for Pr0076 from the same program to verify that our analysis produces consistent results between these different sets of spectra. In our final results we report the stellar parameters and abundances for Pr0076 from both data sets and adopt the results from data taken in 2013. The archive spectra covered a smaller wavelength range (5650–8090 Å) than our new spectra and were obtained in the R = 48,000 mode. One exposure was taken for each star with an integration time of 840 s for Pr0133, 1200 s for Pr0208, 1500 s for Pr0076, and 1500 s for Pr0081. The S/N in the continuum near 6700 Å for these spectra is ∼220.

All of the data were reduced consistently using the MAKEE data reduction software. We required a solar spectrum to derive the solar abundances used to determine the abundances of our target stars relative to the Sun. For this purpose, we used the high-quality Keck/HIRES solar spectrum (S/N ∼ 800 near 6700 Å) from Schuler et al. (2015) obtained in 2010 June in the R = 50,000 mode over the wavelength range 3750–8170 Å. A sample region of the spectra used is shown in Figure 2 for Pr0076 from both data sets in comparison to the solar spectrum.

Zoom In Zoom Out Reset image size

We have derived chemical abundances relative to solar ([X/H]) for up to 20 elements in each of our stars. For our analysis, the adopted solar parameters were T_eff = 5777 K, $\mathrm{log}g=4.44$, and ξ = 1.38 km s⁻¹. A sample of the adopted line lists, EWs, and $\mathrm{log}(N)$ line-by-line abundances for each element is given in Table 2. We derived abundances from measurements of EWs of atomic absorption lines in combination with MOOG, an LTE spectral analysis package (Sneden 1973, version 2014) and ATLAS9 stellar atmosphere models (Kurucz 1993). For the archive spectra, we used an abbreviated line list excluding lines not found in those spectra. To determine the stellar parameters of each star, we require the excitation and ionization balance of Fe i and Fe ii lines. Atomic excitation energies (χ) and transition probabilities ($\mathrm{log}{gf}$) were taken from Mack et al. (2014). For the odd-Z elements Sc, V, Mn, and Co, hyperfine structure (hfs) effects (Prochaska & McWilliam 2000) are taken into account for strong lines through spectral synthesis incorporating hfs components; the resulting abundances are listed in Table 3. The hfs components were obtained from Johnson et al. (2006), and line lists for regions encompassing each feature were taken from VALD (Piskunov et al. 1995; Kupka et al. 1999). Adopted Sc, V, Mn, and Co abundances were derived from the hfs analysis and lines with EWs where hfs effects are negligible. The general method for abundance and error analysis is detailed in Schuler et al. (2011), with new methods used for our current analysis detailed below.

Table 2. Lines Measured, Equivalent Widths, and Abundances

	λ	χ		Solar		Pr0201		Pr0133		Pr0208		Pr0081		Pr0051		Pr0076
Ion	(Å)	(eV)	$\mathrm{log}{gf}$	EW⊙	$\mathrm{log}{N}_{\odot }$	EW	$\mathrm{log}N$	EW	$\mathrm{log}N$	EW	$\mathrm{log}N$	EW	$\mathrm{log}N$	EW	$\mathrm{log}N$	EW	$\mathrm{log}N$
C i	5052.167	7.685	−1.304	31.58	8.417	48.52	8.481	⋯	⋯	⋯	⋯	⋯	⋯	44.66	8.507	⋯	⋯
	5380.337	7.685	−1.615	19.37	8.444	33.17	8.531	⋯	⋯	⋯	⋯	⋯	⋯	30.44	8.566	20.54	8.481
	6587.61	8.537	−1.021	12.2	8.358	28.46	8.565	20.64	8.387	20.58	8.567	14.05	8.454	19.24	8.43	12.22	8.358
	7111.469	8.64	−1.074	10.53	8.431	23.78	8.596	12.28	8.24	15.3	8.544	13.8	8.598	⋯	⋯	10.75	8.443
	7113.179	8.647	−0.762	22.81	8.563	40.39	8.642	30.7	8.482	26.88	8.581	26.74	8.694	30.7	8.561	24.98	8.632
N i	7468.313	10.336	−0.189	4.18	8.131	9.48	8.195	⋯	⋯	⋯	⋯	⋯	⋯	7.64	8.21	⋯	⋯
O i	6300.304	0.0	−9.72	5.10	8.838	4.96	8.930	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	8.44	9.141
	7775.39	9.15	0.001	45.45	8.806	84.57	9.078	84.38	9.139	61.74	9.007	58.58	9.097	67.51	8.951	51.79	8.96
	7774.17	9.15	0.223	64.78	8.925	105.5	9.139	95.89	9.082	77.69	9.039	64.0	8.968	86.72	9.013	64.3	8.96
	7771.94	9.15	0.369	69.65	8.856	119.2	9.16	106.3	9.076	86.34	9.018	72.46	8.956	98.46	9.024	74.15	8.97
Na i	5682.633	2.102	−0.7	98.56	6.282	102.5	6.552	⋯	⋯	116.8	6.555	⋯	⋯	99.82	6.419	119.2	6.524
	6154.226	2.102	−1.56	36.15	6.267	35.74	6.449	35.3	6.359	47.67	6.49	45.52	6.391	37.5	6.403	49.84	6.466
	6160.747	2.104	−1.26	54.69	6.255	49.66	6.374	51.33	6.317	64.53	6.43	61.46	6.321	54.25	6.362	67.11	6.411
Mg i	4730.029	4.346	−2.523	66.13	7.802	60.68	7.927	⋯	⋯	⋯	⋯	⋯	⋯	67.72	7.942	81.46	8.009
	5711.088	4.346	−1.833	103.6	7.597	96.82	7.725	⋯	⋯	117.8	7.827	⋯	⋯	103.6	7.718	119.7	7.798
	6965.409	5.753	−1.51	⋯	⋯	⋯	⋯	⋯	⋯	25.88	7.34	⋯	⋯	⋯	⋯	⋯	⋯
	6841.19	5.753	−1.61	64.3	7.849	69.54	8.004	77.19	8.063	85.91	8.139	⋯	⋯	68.45	7.942	74.84	8.08
Al i	6696.023	3.143	−1.347	37.33	6.253	⋯	⋯	⋯	⋯	47.45	6.453	44.71	6.348	39.17	6.39	51.94	6.47
	6698.673	3.143	−1.647	21.1	6.222	18.94	6.339	20.1	6.292	27.89	6.413	27.36	6.34	22.68	6.366	29.37	6.39
Si i	5701.104	4.93	−1.581	37.85	7.087	39.86	7.236	⋯	⋯	⋯	⋯	⋯	⋯	41.87	7.216	47.47	7.266
	5690.425	4.93	−1.769	50.48	7.485	54.33	7.66	⋯	⋯	59.0	7.641	⋯	⋯	51.69	7.563	57.96	7.621
	5708.4	4.954	−1.034	75.82	7.14	84.46	7.388	⋯	⋯	90.55	7.375	⋯	⋯	80.76	7.268	91.02	7.378
	5772.149	5.082	−1.358	50.81	7.213	58.97	7.445	59.77	7.384	66.03	7.466	62.82	7.402	56.25	7.35	63.21	7.42
	7405.772	5.614	−0.313	89.3	7.108	100.7	7.352	102.1	7.332	104.6	7.326	106.4	7.324	100.7	7.289	106.3	7.34
	6125.021	5.614	−1.464	32.55	7.48	39.8	7.688	39.12	7.62	41.65	7.658	39.71	7.614	38.83	7.631	41.96	7.657
	6244.466	5.616	−1.093	44.9	7.316	54.31	7.54	53.57	7.477	58.57	7.539	55.23	7.479	53.51	7.484	56.47	7.505
	6243.815	5.616	−1.242	44.62	7.46	53.32	7.675	55.52	7.655	57.63	7.675	53.49	7.604	48.09	7.555	62.3	7.734
	6145.016	5.616	−1.31	39.69	7.45	46.7	7.644	42.13	7.518	53.6	7.685	43.08	7.516	44.76	7.572	48.16	7.6
	6142.483	5.619	−1.295	33.85	7.339	39.73	7.522	35.73	7.397	41.73	7.495	39.28	7.443	38.38	7.459	43.42	7.516
	6848.58	5.863	−1.524	17.25	7.401	21.13	7.572	19.82	7.485	23.61	7.584	23.82	7.582	19.27	7.491	24.55	7.602
	6414.98	5.871	−1.035	46.89	7.484	55.2	7.671	⋯	⋯	⋯	⋯	⋯	⋯	54.52	7.622	59.12	7.665
	7003.569	5.964	−0.937	57.28	7.606	61.99	7.74	68.65	7.794	77.06	7.873	66.17	7.734	59.56	7.669	62.59	7.697
	6741.628	5.984	−1.428	15.06	7.344	20.56	7.56	19.67	7.49	22.05	7.553	19.79	7.491	18.54	7.477	20.47	7.511
S i	4694.113	6.525	−1.77	11.0	7.311	13.81	7.27	⋯	⋯	⋯	⋯	⋯	⋯	16.24	7.419	13.26	7.423
	4695.443	6.525	−1.92	6.13	7.167	11.15	7.307	⋯	⋯	⋯	⋯	⋯	⋯	9.21	7.267	9.29	7.382
	6757.171	7.87	−0.31	14.14	7.197	34.72	7.514	28.69	7.393	23.76	7.438	⋯	⋯	25.91	7.402	19.24	7.394

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

Table 3. Synthesized HFS Line Abundances

	λ	χ		Solar
Ion	(Å)	(eV)	$\mathrm{log}{gf}$	$\mathrm{log}{N}_{\odot }$	Pr0201	Pr0133	Pr0208	Pr0081	Pr0051	Pr0076
Co i	6814.942	1.956	−1.9	⋯	5.07	⋯	⋯	⋯	⋯	⋯
	5301.039	1.71	−2.0	⋯	⋯	⋯	⋯	⋯	⋯	⋯
Mn i	5432.546	0.0	−3.795	5.28	5.46	⋯	⋯	⋯	5.41	5.55
Sc ii	6604.601	1.357	−1.309	3.10	3.18	3.09	3.25	3.15	3.17	3.27
	6245.637	1.507	−1.03	3.10	3.23	3.09	3.25	3.15	3.16	3.17
V i	6111.645	1.043	−0.715	⋯	⋯	⋯	⋯	⋯	⋯	4.05
	6090.214	1.081	−0.062	3.83	4.03	⋯	⋯	⋯	⋯	4.05

Note. Listed synthesized abundances replace EW abundances from Table 2.

Download table as:  ASCIITypeset image

The EW measurement of absorption lines is performed by an in-house user-guided code called eXtract from SPECTra—Equivalent Widths (XSpect-EW), specifically created for this purpose. The general process for measuring EWs of absorption lines for abundance derivations includes three main steps: normalization of the continuum, wavelength shift, and fitting the absorption lines with Gaussian or Voigt profiles to determine the EW of the lines. Each step of the process can be vulnerable to user error if done manually, as detailed in the following subsections, depending on how much care and time is given to each part of the analysis. Ideally, one would take extreme care in each of these parts within a minimal amount of time.

2.4.1. Continuum Normalization

Arguably the most important part of the EW measuring process is the ability to consistently determine the continuum of the spectrum or each spectral order. The differential curve-of-growth (CoG) abundance determination involves comparing the abundances of the star of interest to those of a standard or reference star, often the Sun; therefore, one must be able to determine the continuum in each spectral order in the same way for the star and the standard so as to minimize differences in the abundances that could arise as a result of the analysis. XSpect-EW has been designed to normalize the spectrum efficiently and accurately by using a two-step process.

Step 1: a spectrum is split into smaller pieces (referred to as selection windows), the size of which is set to larger than the typical size of the absorption lines to ensure that the selection windows do not fall entirely within a line. For example, in an optical high-resolution Keck/HIRES spectrum with R > 60,000, the width of a typical line will be 0.4–0.6 Å, so the size of the selection window could be ∼1.6 Å. Points above 90% of the flux values within a selection window are chosen as points in the continuum (90% value can be adjusted by the user if needed).

Step 2: the chosen continuum points are used to normalize the spectrum by fitting them with a Gaussian process (GP). A GP is a collection of random variables, any finite number of which have a joint Gaussian distribution. GPs are useful for data containing nontrivial stochastic signals or noise. A square exponential kernel (or Gaussian kernel) ${k}_{\mathrm{SE}}{(x,x^{\prime} )=A\exp [-{\rm{\Gamma }}(x-x^{\prime} )}^{2}]$ is used for the covariance function in the GP, where A is the variance and Γ is the inverse length scale (Γ = 1/2l²). A variance of similar order of magnitude to the flux values is used (10⁷), along with an inverse length scale of 0.1. After testing various values, these gave the best results for our spectra. XSpect-EW will be updated to automatically determine the values of variance and length scale for each order in a spectrum and to have a variety of kernels available. The flux is then divided by the curve output by the GP (gray curve in Figure 3 top panel), which normalizes the order (Figure 3 bottom panel).

Zoom In Zoom Out Reset image size

2.4.2. Wavelength Shift

2.4.3. Equivalent Width Measurements

Measuring the EW of an absorption line generally involves fitting a Gaussian or Voigt function to the line and then integrating that function to determine the area enclosed by the curve and continuum. XSpect-EW uses a four-step process for this.

Step 1: the extent of the line being measured is determined. Carefully determining the extent, or width, of the line at the continuum, is critical to obtaining an accurate measure of the absorption in the line. A piece of the spectrum around a selected line is trimmed from the rest of the spectrum (default set to 1.5 Å about the line center). The boundaries of the line being measured are determined by utilizing the slope of the flux values within the piece of the spectrum. These values range from a maximum or minimum near the center of the line to zero at the continuum and the core, as shown in Figure 4. Using one half of the standard deviation of the flux slope values, a range of values centered on the continuum is created. After testing various values, one half of the standard deviation gave us best results relative to hand-measured EWs. The boundaries of the absorption line are defined as the points where the slope enters this range from the maximum or minimum values to either edge of the line. At this boundary, the slope values approach zero as the flux values approach the constant continuum.

Zoom In Zoom Out Reset image size

Step 2: with the boundaries defined, XSpect-EW assumes the continuum is correct and shifts the data outside the boundaries of the line (still within the 1.5 Å line window) to match the continuum, effectively isolating the line to be measured.

Step 3: the isolated absorption line and associated continuum are fed into another GP (same kernel) with a small variance of 1 and large inverse length scale of 100.

Step 4: the GP can produce different realizations of the data, and XSpect-EW uses these to perform a Gaussian fit to the line, repeating the process 100 times and thus producing 100 different EW measurements. The mean and standard deviation are calculated using the 100 measurements of the line. Each absorption line can be plotted and remeasured if needed, with the ability to adjust the local continuum and the center and boundaries of the line.

2.4.4. Verification of Methodology

To verify the robustness of our new EW measuring tool, in Figure 5 we compare EWs of Fe i and Fe ii lines measured in our solar spectrum with XSpect-EW and by hand with SPECTRE, a spectral analysis package (Sneden et al. 2012). All differences are <15%, with 80% of Fe lines having differences smaller than 5%. In the remaining elemental lines (not shown here), ∼70% of lines have differences smaller than 5%. XSpect-EW requires some user input to obtain these measurements, which consists of checking each automatic line fit and remeasuring problematic lines by adjusting extra parameters that characterize the absorption line. We note that this interactive functionality has been built into XSpect-EW for this specific purpose. The scatter in the fractional difference of the line measurements is larger for the other elements, because those lines can generally be more difficult to measure due to the variety in the strength of the lines and proximity to other lines. A well curated line list can be helpful for automatic line fits because strong, isolated absorption lines are easier for XSpect-EW to measure automatically. The quality of the data will also be a deciding factor in the number of lines that need to be remeasured. In this example, 17% of Fe lines and 23% of other elemental lines required remeasuring, greatly reducing the amount of time needed to measure all lines and allowing the user to focus on problematic lines to produce abundances with high precision. Lines that are remeasured are only considered based on visual inspection of the fit, not in comparison to previous measurements, since no comparison would be available when measuring a new star. The average of the Fe absolute abundances derived from the XSpect-EW and SPECTRE EWs, shown in Figure 6, agrees within errors and agrees with the input solar absolute abundance within MOOG of 7.50 dex. For all Fe i and ii lines, we see a smaller scatter in the absolute abundances from the XSpect-EW measurements than in the SPECTRE measurements by 0.01 dex. From this we can see that XSpect-EW performs as well as, if not slightly better than, a full set of hand-measured EWs from SPECTRE.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

From the Fe i and Fe ii abundances derived using the EWs measured with XSpect-EW, we determine an average cluster metallicity for Praesepe of 0.21 ± 0.02 dex, which is in good agreement with most past works on the cluster metallicity. In Table 4, we show the stellar parameter values in the literature for each star along with our own derived values. Our derived stellar parameters (${T}_{\mathrm{eff}},\mathrm{log}G,[\mathrm{Fe}/{\rm{H}}]$) are also in good agreement with current literature values. The main discrepancy between our stellar parameters and literature values comes from the ξ parameter values, which in general are larger than literature values, likely due to the higher ξ value adopted for the Sun in this work (1.38 km s⁻¹ here compared to lower values used in other studies) (Pace et al. 2008; Gebran et al. 2019; D'Orazi et al. 2020). A more detailed comparison of individual stars to the literature is presented in Section 3.2.

Table 4. Comparison to Literature

Star ID	Teff	${\rm{log}}(G/{\rm{g}}\,{\rm{cm}}\,{{\rm{s}}}^{-1})$	ξ	[Fe/H]	References
alt ID	(K)		(km s−1)	(dex)
Pr0201	6168 ± 35	4.34 ± 0.10	1.52 ± 0.06	0.23 ± 0.05	This work
Prae kw 418	6174 ± 50	4.41 ± 0.10	⋯	0.19 ± 0.04	Quinn et al. (2012)
	6062 ± 110	4.44 ± 0.07	1.27 ± 0.18	0.24 ± 0.10	Pace et al. (2008)
Pr0133	6067 ± 60	4.18 ± 0.12	1.74 ± 0.11	0.19 ± 0.06	This work
Prae kw 208	6005 ± 19	4.46 ± 0.21	1.05 ± 0.04	0.18 ± 0.03	Gebran et al. (2019)
	5997 ± 60	4.38 ± 0.20	1.40 ± 0.20	0.12 ± 0.10	Boesgaard et al. (2013)
	5993 ± 110	4.45 ± 0.07	1.52 ± 0.18	0.28 ± 0.10	Pace et al. (2008)
Pr0208	5869 ± 46	4.37 ± 0.13	1.53 ± 0.07	0.26 ± 0.07	This work
N2632-8	5977 ± 75	4.55 ± 0.15	1.30 ± 0.20	0.25 ± 0.11	D'Orazi et al. (2020)
Prae kw 432	5841 ± 73	4.40 ± 0.20	1.25 ± 0.20	0.17 ± 0.10	Boesgaard et al. (2013)
Pr0081	5731 ± 42	4.44 ± 0.11	1.41 ± 0.06	0.18 ± 0.06	This work
CPrae kw 30	5716 ± 45	4.57 ± 0.42	1.18 ± 0.04	0.12 ± 0.04	Gebran et al. (2019)
	5675 ± 111	4.44 ± 0.20	1.07 ± 0.20	0.12 ± 0.10	Boesgaard et al. (2013)
Pr0051	6017 ± 27	4.40 ± 0.07	1.54 ± 0.05	0.16 ± 0.03	This work
TYC 1395-668-1
Pr0076	5748 ± 24	4.44 ± 0.07	1.35 ± 0.03	0.22 ± 0.05	This work
Prae kw 23	5773 ± 53	4.56 ± 0.24	1.20 ± 0.04	0.20 ± 0.04	Gebran et al. (2019)
	5699 ± 79	4.43 ± 0.20	1.10 ± 0.20	0.12 ± 0.10	Boesgaard et al. (2013)

Download table as:  ASCIITypeset image

The derived elemental abundances for our targets are summarized in Table 5. Elements with only one measured absorption line have no mean and therefore no uncertainty in the mean; for those lines we have adopted the total error to be the average total error of all other elements in the star. Some elements in some stars were not measurable due to lack of spectral coverage or removal of lines that were not measurable (due to noise, blending, bad fit, etc.). The errors for all stellar parameters and abundances are symmetric or close to symmetric (where ±σ_Total intervals are equal or close to equal); in all cases we conservatively adopt the larger error, except with surface gravity where we adopt the average error. The resulting differences between the archive data (2003) and our data (2013) for Pr0076 are within error bars for the derived stellar parameters and elemental abundances, as shown in Table 5. Despite the fact that we are using data of differing quality, observing conditions, instrument setup, and time of observation, we are able to produce consistent results, giving us great confidence in our analysis.

Table 5. Stellar Abundances

	Pr0201	Pr0133	Pr0208	Pr0081	Pr0051	Pr0076 (2003)	Pr0076 (2013)
[C/H]	0.12 ± 0.03 a ± 0.05 b	0.02 ± 0.03 ± 0.06	0.12 ± 0.06 ± 0.09	0.13 ± 0.03 ± 0.05	0.07 ± 0.03 ± 0.04	0.10 ± 0.07 ± 0.09	0.03 ± 0.01 ± 0.04
[O/H]	0.11 ± 0.03 ± 0.05	0.07 ± 0.03 ± 0.07	0.09 ± 0.05 ± 0.08	0.10 ± 0.09 ± 0.11	0.05 ± 0.03 ± 0.05	⋯	0.07 ± 0.04 ± 0.05
[Na/H]	0.19 ± 0.02 ± 0.03	0.12 ± 0.04 ± 0.05	0.22 ± 0.04 ± 0.03	0.09 ± 0.03 ± 0.04	0.13 ± 0.01 ± 0.02	0.27±... ±0.08	0.20 ± 0.03 ± 0.03
[Mg/H]	0.14 ± 0.01 ± 0.03	0.18±... ±0.05	0.23 ± 0.03 ± 0.05	⋯	0.12 ± 0.01 ± 0.02	0.20 ± 0.24 ± 0.24	0.21 ± 0.01 ± 0.05
[Al/H]	0.12±...±0.05	0.11±... ±0.05	0.20 ± 0.00 ± 0.02	0.11 ± 0.01 ± 0.02	0.14 ± 0.00 ± 0.01	⋯	0.19 ± 0.02 ± 0.03
[Si/H]	0.20 ± 0.01 ± 0.01	0.15 ± 0.01 ± 0.02	0.20 ± 0.01 ± 0.01	0.14 ± 0.01 ± 0.01	0.12 ± 0.01 ± 0.01	0.19 ± 0.02 ± 0.02	0.18 ± 0.01 ± 0.01
[Ca/H]	0.19 ± 0.02 ± 0.03	0.17 ± 0.01 ± 0.05	0.23 ± 0.02 ± 0.05	0.17 ± 0.02 ± 0.04	0.16 ± 0.01 ± 0.03	0.27±... ±0.08	0.24 ± 0.01 ± 0.03
[Sc/H]	0.12 ± 0.02 ± 0.05	−0.01±... ±0.05	0.16 ± 0.01 ± 0.07	0.05±... ±0.05	0.06 ± 0.01 ± 0.03	⋯	0.12 ± 0.05 ± 0.07
[Ti/H]	0.22±0.01 ± 0.06	0.13 ± 0.01 ± 0.06	0.24 ± 0.01 ± 0.04	0.19 ± 0.03 ± 0.05	0.15 ± 0.01 ± 0.04	0.24 ± 0.03 ± 0.10	0.22 ± 0.02 ± 0.06
[V/H]	0.22 ± 0.03 ± 0.05	0.28 ± 0.05 ± 0.08	0.25 ± 0.02 ± 0.06	0.19 ± 0.02 ± 0.05	0.16 ± 0.01 ± 0.03	0.27 ± 0.03 ± 0.08	0.24 ± 0.01 ± 0.03
[Cr/H]	0.20 ± 0.01 ± 0.03	0.21 ± 0.01 ± 0.05	0.27 ± 0.03 ± 0.05	0.21 ± 0.02 ± 0.04	0.19 ± 0.01 ± 0.02	0.29 ± 0.01 ± 0.05	0.26 ± 0.01 ± 0.02
[Mn/H]	0.20±... ±0.04	⋯	⋯	⋯	0.13±... ±0.03	0.36 ± 0.01 ± 0.07	0.27±... ±0.04
[Fe/H]	0.23 ± 0.01 ± 0.05	0.19 ± 0.01 ± 0.06	0.26 ± 0.01 ± 0.07	0.18 ± 0.01 ± 0.06	0.16 ± 0.00 ± 0.03	0.25 ± 0.01 ± 0.06	0.22 ± 0.00 ± 0.05
[Co/H]	0.11 ± 0.01 ± 0.03	0.14±... ±0.05	0.27 ± 0.02 ± 0.04	0.14 ± 0.01 ± 0.03	0.10 ± 0.03 ± 0.04	0.18 ± 0.03 ± 0.07	0.16 ± 0.02 ± 0.03
[Ni/H]	0.19 ± 0.01 ± 0.02	0.14 ± 0.01 ± 0.04	0.21 ± 0.01 ± 0.03	0.14 ± 0.01 ± 0.03	0.13 ± 0.01 ± 0.02	0.24 ± 0.01 ± 0.05	0.21 ± 0.01 ± 0.02
[Zn/H]	0.15±...±0.05	⋯	⋯	⋯	0.12±... ±0.03	0.18±... ±0.08	0.15±... ±0.04
[S/H]	0.23 ± 0.09 ± 0.09	0.20±... ±0.05	0.20±... ±0.05	⋯	0.12 ± 0.02 ± 0.03	0.22 ± 0.02 ± 0.05	0.20 ± 0.03 ± 0.05
[Cu/H]	0.02±...±0.05	−0.09±... ±0.05	0.14±... ±0.05	0.18±... ±0.05	−0.02±... ±0.03	0.28±... ±0.08	0.22±... ±0.04
[K/H]	0.34±...±0.05	0.31±... ±0.05	⋯	0.12±... ±0.05	0.29±... ±0.03	⋯	0.18±... ±0.04
[N/H]	0.06±...±0.05	⋯	⋯	⋯	0.08±... ±0.03	⋯	⋯

Notes.

^a σ_μ —the uncertainty in the mean. ^b σ_Total—quadratic sum of σ_μ and uncertainties due to T_eff, $\mathrm{log}g$, and ξ.

Download table as:  ASCIITypeset image

Figure 7 shows the abundance versus atomic number for each star and the cluster mean. Most elements have a similar spread about the mean while some (K, Sc, Co, and Cu) show a larger spread. This larger spread is likely due to the fact that these elements have a low number of lines measured, lines may be weak, noisy, or blended with other nearby lines and thus difficult to measure, or some combination of these factors.

Zoom In Zoom Out Reset image size

3.1.1. Planet-hosting Star

In order to see whether the planet host, Pr0201, is different from the rest of our sample, we compare the abundances to the cluster mean. Figure 8 shows the difference between Pr0201 and the cluster mean (not including the planet host) for each derived element versus atomic number. Most elements fall between ±0.05 dex of the cluster mean and are within errors of zero with a few exceptions. The element with the largest difference is K, having an overabundance of 0.12 dex when compared to the cluster. Si, having the smallest total error of 0.01 dex, is also significantly overabundant in the planet host by 0.04 dex. The planet host shows no obvious trend in abundance versus atomic number over all measured elements.

Zoom In Zoom Out Reset image size

3.1.2. Non-planet-hosting Stars

In Figure 9 we show the difference between each star and the cluster mean (the mean not including that specific star). Elements noted below are considered overabundant or deficient if they are more than 1σ (in actual errors, not average errors) different from zero.

Zoom In Zoom Out Reset image size

Pr0133 is only overabundant in K by 0.08 dex and deficient in C by 0.07 dex, Ti by 0.07 dex, Sc by 0.11 dex, and Cu by 0.20 dex.

Pr0208 has the highest metallicity of our analyzed stars, [Fe/H] of 0.26 dex, and shows an overabundance in many elements: S by 0.04 dex, Ni by 0.05 dex, Ti by 0.06 dex, Cr by 0.06 dex, Na by 0.07 dex, Mg by 0.07 dex, Al by 0.07 dex, Sc by 0.09 dex, and Co by 0.14 dex. While elements of high condensation temperature tend to be overabundant in Pr0208, the large error bars on the volatile elements make it difficult to say whether there is any overall trend with condensation temperature.

Pr0081 shows an overabundance in C by 0.06 dex and Cu by 0.13 dex. Deficient elements include: Si by 0.03 dex, Al by 0.04 dex, Na by 0.08 dex, and K by 0.16 dex.

Pr0051 shows no overabundance but many deficiencies: Cr by 0.04 dex, Si by 0.05 dex, Ni by 0.05 dex, Ti by 0.05 dex, V by 0.08 dex, Fe by 0.06 dex, Mg by 0.07 dex, Co by 0.06 dex, S by 0.09 dex, Mn by 0.09 dex, and Cu by 0.11 dex.

Pr0076 shows deficiencies in C by 0.06 dex and K by 0.08 dex while being overabundant in Cr by 0.04 dex, Na by 0.05 dex, Al by 0.05 dex, Ni by 0.05 dex, Ca by 0.06 dex, Mn by 0.11 dex, and Cu by 0.17 dex.

An et al. (2007) determined [Fe/H] of four G-type dwarfs in Praesepe through spectroscopy (equivalent width method) and photometry (photometric metallicity). The mean [Fe/H] they determined through spectroscopy (+0.11 ± 0.03) was lower than our determined mean metallicity of 0.21 ± 0.02. The [Fe/H] they determined through photometry (+0.20 ± 0.04) is in much better agreement with our results and most current literature values.

Pace et al. (2008) performed a detailed chemical study of eight elements (Fe, Na, Al, Si, Ca, Ti, Cr, Ni) on 20 solar-type stars in four open clusters, obtaining data with high resolution (R = 100,000) and high signal-to-noise ratio (S/N = 130) for seven stars in Praesepe (with two overlapping stars: Pr0133 and Pr0201). They measure a higher mean Fe abundance of +0.27 ± 0.10, though still within errors of our mean abundance. For Pr0201, all eight elements agree within errors, while Pr0133 shows less agreement. Half of the elements agree with our results (Ca, Ti, Cr, Fe) while the other half (Na, Al, Si, Ni) are higher in abundance than our measurements. Our Al measurement comes from just one absorption line for Pr0133, with the total error being the average of the other elements; this could mean this error is underestimated. It is unclear why there is such a discrepancy in the other elements. All reported mean abundances relative to Fe are within error bars except for O, which is much lower than our determined abundance.

Carrera & Pancino (2011) studied abundances of three giants in Praesepe using the equivalent width method, and while they measure a lower mean [Fe/H] abundance of +0.16 ± 0.05 dex, this mean does fall within errors of our determined mean. Many of the other abundances relative to Fe (Al, Ca, Co, Cr, Mg, Na, Ni, Sc, Si, Ti, and V) disagree with our results, having higher abundances except in the case of Ti and Ca. This is likely due to the fact that these stars are of different stellar type than ours and the lower measured metallicity would increase the abundance ratios relative to Fe.

Boesgaard et al. (2013) presented chemical abundances of 16 elements (Li, C, O, Na, Mg, Al, Si, Ca, Sc, Ti, V, Cr, Fe, Ni, Y, and Ba) for 11 solar-type stars in Praesepe (with four overlapping stars: Pr0133, Pr0208, Pr0081, and Pr0076) through equivalent width analysis. They determined a mean cluster metallicity of +0.12 ± 0.04 dex, which is lower than our determined mean by 0.09 dex. Abundance ratios for other elements relative to Fe are within 1σ of our measurements except for Al and Sc. In most cases the abundance ratios are higher in value, likely because of the lower mean [Fe/H] abundance. In Pr0133, Pr0208, and Pr0076, our study overlaps with 13 elements (excluding Li, Y, and Ba) while Pr0081 overlaps with 12 elements (excluding Mg as well). For Pr0133, all elements agree within errors except for Sc, which was found to be about solar or higher (0.04 ± 0.08 dex) while we derived a subsolar value (−0.20 ± 0.08 dex). For Pr0208, errors are not reported in this study, but most elements agree within our own errors. Two elements that do not agree within our errors are Sc and Ti; however, if the errors on these elements are at all similar to the errors for the other stars reported in Boesgaard et al. (2013) then it is very likely they would agree with our abundances. Abundances for Pr0081 are also presented without errors, and similarly to Pr0208, all derived abundances (Fe, C, O, Na, Al, Si, Ca, Sc, Ti, V, Cr, and Ni) are within our own errors except for Sc. It is also true that if the errors for this star are similar to the other stars, Sc would also fall within the errors. For Pr0076, all 13 elements fall within reported errors.

Gebran et al. (2019) used an automated spectral analysis code (BACCHUS) to determine abundances of 24 elements for five stars in Praesepe (with three overlapping stars: Pr0133, Pr0081, and Pr0076). They report mean abundances for G and K stars separately, +0.17 ± 0.04 and +0.12 ± 0.01 dex respectively. These means are consistent with each other and the literature; the average for G types is consistent with our work. In Pr0133 and Pr0081, our study overlaps with 14 elements (C, O, Na, Mg, Al, Si, Ca, Sc, Ti, V, Cr, Co, Ni, and Cu), all of which agree within errors for Pr0081. For Pr0133, all elements but Na and Cu agree within errors. Our [Na/Fe] abundance is about solar or less (−0.07 ± 0.07 dex) while Gebran et al. (2019) determined an even lower abundance of −0.35 ± 0.16 dex. The reverse is true for Cu: our work derived a very low abundance of −0.28 ± 0.09 dex while Gebran et al. (2019) derived a solar or higher value of 0.02 ± 0.03 dex. For Pr0076, our study overlaps with 17 elements (Fe, C, O, Na, Mg, Al, Si, Ca, Sc, Ti, V, Cr, Mn, Co, Ni, Cu, and Zn), all of which also fall within reported errors.

D'Orazi et al. (2020) revisited the metallicity of Praesepe, taking high-resolution spectroscopic observations of 10 solar-type dwarf stars, including Pr0208. All eight elements studied (Fe, Na, Mg, Al, Si, Ca, Ti, and Ni) are within errors of our derived abundances for Pr0208. They report a mean metallicity of +0.21 ± 0.01 dex and conclude that Praesepe is the most metal-rich, young open cluster in the solar neighborhood, in remarkable agreement with our results, which also gives confidence in works that report a higher metallicity. Pr0051 has no literature values to be compared to but the general agreement of our results to the literature gives us confidence in those values as well.

Even with similar analyses the literature suggests a large scatter in metallicity for Praesepe; however, each individual work shows a much lower scatter within its respective sample. The scatter within these works (for [X/Fe]), including our own, is largely consistent with what we would expect from open clusters (De Silva et al. 2007; Bovy 2016; Kovalev et al. 2019). This implies that the Praesepe cluster is chemically homogeneous and these works may be self-consistent but they may not all be consistent with each other. As our analyses and precision improve, it seems we may converge on a consistent metallicity for Praesepe. Perhaps a larger sample analyzed consistently could shed more light on the mean and scatter for the cluster.

In Table 6 we show the mean, standard deviation, and error in the mean of [X/H] and [X/Fe] abundance for our stars in Praesepe. For most of the elements, the standard deviation and the error in the mean of [X/Fe] are lower than either for [X/H]. A few exceptions are the elements C, N, O, and Zn, which show a lower standard deviation and error in the mean of [X/H]. K shows the same standard deviation and error in the mean in both [X/H] and [X/Fe]. The average standard deviation is ∼0.01 dex lower in [X/Fe] than in [X/H], while the average error in the mean is about the same. In [X/Fe], the standard deviation of most elements is at or below the average of ∼0.02–0.03 dex, well within literature limits on abundance scatter in open clusters (De Silva et al. 2007; Bovy 2016; Kovalev et al. 2019). With this, we verify that the stars in our sample are chemically homogeneous; a larger sample would be required to confirm the chemical homogeneity of Praesepe as a whole. C and N are above the average by ∼0.02 dex. K and Cu are much higher than the average by ∼0.04 and ∼0.06 dex respectively. Looking at the error in the mean, we see a similar separation where most elements are below the average at ∼0.01–0.02 dex, with elements N, K, and Cu being ∼0.03 dex higher than the average. C and Mn are above the mean by ∼0.01 dex or less.

For elements with Z > 19 (Ca, Sc, Ti, V, Cr, Mn, Co, Ni, and Zn), the scatter in both standard deviation and error in the mean is higher for odd-Z elements (Sc, V, Mn) with the exception of Co, which is similar to Ca, Ti, and Cr, but higher than Ni and Zn. This is likely due to the fact that some of these odd-Z elements may have one or two synthetically measured absorption lines for each star to account for hfs effects, along with lines measured normally where hfs effects are negligible.

In the search for signatures of planet formation we specifically focus on trends in refractory elements with T_C > 900 K (Meléndez et al. 2009). During early disk evolution, elements with high T_C are expected to condense into solids at shorter distances from the host star, leading to refractory-poor gas and refractory-rich planetesimals. This results in two possible abundance signatures for the host star. The removal of these elements from the protoplanetary disk allows the accretion of the refractory-depleted material onto the host star, which imparts a decreasing trend in the refractory elements with T_C. Alternatively, accretion of refractory-rich planetesimals or planets themselves would impart an increasing trend.

In this section we interpret trends in condensation temperature in the context of a planet engulfment model, explained in Mack et al. (2014), based on the addition or removal of material with similar composition to the Earth to/from the convection zone of a solar-type star. The composition of the Earth is taken from McDonough (2001) and solar composition is from Asplund et al. (2009). We adjust the size of the convection zone based on the temperature of the star according to Pinsonneault et al. (2001). Solar abundances are modified to match those of the cluster mean (excluding Pr0201), then we can adjust the number of M_⊕ accreted/sequestered in order to produce a desired T_C slope.

Figure 10 shows the T_C trend for refractory elements of the cluster mean. Here, we have excluded the planet host, Pr0201, to show that the mean cluster abundances alone present no trend with condensation temperature. The cluster mean abundances show a slope of −5.98 × 10⁻⁶ ± 3.25 × 10⁻⁵ dex K⁻¹, consistent with zero slope. When comparing each other star with the cluster mean (excluding that star and the planet host from the mean), no star shows a statistically significant trend in refractory elements with condensation temperature (slopes shown in Figure 11).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We will take a closer look at Pr0201, because it is the only known planet host in our sample. Figure 12 shows the T_C trend for refractory elements in Pr0201 relative to the cluster mean. The best-fit line to these elements gives a negative slope of −8.64 × 10⁻⁵ ± 6.59 × 10⁻⁵ dex K⁻¹. While not a statistically significant detection (1.3σ), if taken at face value, this trend could be explained by the sequestering of ∼1.62 M_⊕ of material from the convection zone of Pr0201.

Zoom In Zoom Out Reset image size

Our results for T_C slope are relative to solar abundances ([X/H]), meaning that a slope consistent with zero tells us the distribution of abundances with respect to T_C is similar to that of the Sun. Results for the cluster mean T_C slope are consistent with zero slope, meaning that the abundance distribution of the cluster may be similar to that of the Sun. This could indicate that planet formation for Sun-like stars in the cluster is common or that the cluster initially formed from a cloud with this distribution already in place. The possible discovery of more planets within the cluster could give more weight to the prominent planet-forming case. All six of our analyzed stars, including the known planet host, show a slope consistent with zero when compared to the cluster mean. If planet formation is in fact prevalent, the lack of detected planets in our sample could be due to the difficulty of finding smaller planets or these systems may have lost their planets entirely.

The Pr0201 system is host to a gas giant (Pr0201b) with a minimum mass of 0.54 ± 0.039 M_J in a circular orbit (Quinn et al. 2012) of short period (4.4264 ± 0.0070 days). With such a close-in giant planet and a T_C slope consistent with zero, we investigate the possibility that Pr0201 could have accreted or sequestered refractory-rich material during the planet formation process. Using our planet engulfment model mentioned above, we can place limits on the amount of this material. To do this, we assume Pr0201 formed with the same composition as the cluster mean, starting off with zero slope in the refractory elements. We then add material with the same composition as the Earth until we produce a significant T_C slope (3σ using the measured error in the T_C slope for Pr0201) in the refractory elements (shown in the top panel of Figure 13). In order to produce a statistically significant T_C slope, Pr0201 would have needed to accrete 4.42 M_⊕ of material. In the bottom panel of Figure 13, we show how this would affect the individual refractory elements of the cluster mean, depicted by the orange points and lines, which can be compared to the abundances of Pr0201 in purple. The solid lines show the mean abundance while the dashed lines show ± the error in the mean. This accretion scenario would produce abundances that are noticeably enhanced compared to Pr0201 and can be ruled out at the ∼16.5σ level. Excluding Ti, V, and Fe, it is also the case that 1.25 M_⊕ of material can be ruled out at the 5σ level, indicated by the green lines. In general, Pr0201 does not seem significantly enhanced in refractory elements when compared to the cluster mean. This analysis focuses on accretion but can also be applied to material sequestered.

Zoom In Zoom Out Reset image size