UTILIZING SYNTHETIC UV SPECTRA TO EXPLORE THE PHYSICAL BASIS FOR THE CLASSIFICATION OF LAMBDA BOÖTIS STARS

Since Morgan et al. (1943) gave the first definition of Lambda Boo stars, Lambda Boo candidates published in the literature have been selected using several different criteria, and sometimes the agreement between the resulting lists is quite poor. Until all the classification criteria converge into a quantitative set of class characteristics, we will not be able to discriminate between theories explaining the origin of the Lambda Boo phenomenon.

We started with a list of over 200 stars cited in the literature as Lambda Boo stars. This list led to the compilation of possible Lambda Boo stars presented in Murphy et al. (2015). In this study, we carried out a more detailed evaluation of the ultraviolet properties of "confirmed" Lambda Boo stars in order to select our final sample. From this, we selected 33 unambiguous Lambda Boo stars that have archival ultraviolet spectra (Table 1) to form the basis of the present study. All of the stars selected for this study, except for HD 101108, are classified as "confirmed" Lambda Boo stars in Murphy et al. (2015).

Because we are exploring classification criteria for the Lambda Boo class as a function of T_eff and [M/H], we have searched for all of the published T_eff and [M/H] values for each star in our study. Since most of our program stars have more than one effective temperature cited in the literature, we needed a mean effective temperature for the purpose of our study. We assigned a mean effective temperature to each star based on a range of temperatures found in the literature (excluding any outliers) and we did not include any duplicate temperatures. We also limited the number of sources included in this range to the five values that best represented the data set for each star. The mean effective temperature, a range of effective temperatures, and the observations we used in this paper are listed in Tables 1–3. We also list previously published metallicity values, which allow us to compare our program stars to models with different metallicities. The methods that have been used to determine metallicity differ from paper to paper, and sometimes [M/H] and [Fe/H] are used interchangeably. We cite them as [M/H] or [Fe/H] based on how they are labeled in their original publications.

We have compared our Lambda Boo stars with a set of normal (standard) stars selected from the "IUE Ultraviolet Spectral Atlas of Standard Stars"⁶ and various sources of MK standards (e.g., Mamajek's website⁷ and Jaschek & Gomez 1998). The normal stars used in this study can be found in Table 2.

Since Lambda Boötis itself and most of our other program Lambda Boo stars have never been observed with the Hubble Space Telescope (HST), we can only use archival IUE spectra for this study. More than 45% of the known Lambda Boo stars (Murphy et al. 2015) were observed with the IUE during its 18 year operation period. The spectral resolution achievable with IUE is limited by the size of the point-spread function in the camera image plane. In general, this varies with wavelength and is dependent on the camera used, the dispersion mode, and camera focusing conditions. IUE performed spectrophotometry at high (0.1–0.3 Å) and low (6–7 Å) resolutions between 1145 and 3300 Å. The SWP camera covers 1145–1975 Å and the long-wavelength LWP/LWR cameras cover 1850–3300 Å. The specific IUE spectra we used for our own analysis are listed in Tables 1 and 2. They were all acquired from the data archive hosted by the North American Mikulski Archive for Space Telescopes (MAST)⁸ and processed from the NEWSIPS data processing systems (Nichols & Linsky 1996).

We also examined archival visible spectra of our program stars (Table 3) to determine whether a weak Mg ii line is an exclusive signature of Lambda Boo stars (see Section 4.4). We obtained the visible spectra from the ELODIE archive, which contains the complete collection of high-resolution echelle spectra accumulated using the ELODIE spectrograph at the Observatoire de Haute-Provence 1.93 m telescope (Moultaka et al. 2004). These spectra have a nominal resolution of R  =  42,000.

The UV criteria for Lambda Boo stars are largely based on absorption line ratios. We have measured the equivalent widths of the C i 1657 Å, Al ii 1671 Å, and Mg ii 2800 Å absorption lines (Table 4) in IUE spectra of all of our confirmed Lambda Boo stars and comparison standards. We also used the same procedure to measure the equivalent widths of these lines in every synthetic spectrum we generated (our modeling procedures are described in the next section). In order to compare the equivalent width measurements of observed and synthetic spectra we needed a repeatable way to both determine the shape of the local continuum and to integrate over consistent wavelength ranges. We have shifted all observed wavelengths to lab values using radial velocities found in the literature. The program used to perform our equivalent width measurements, "autocfeature," is a modified version of the program "feature" in the IUEDAC IDL library.⁹

We modified the original "feature" program to include spline fitting and auto-wavelength selection capabilities for the purpose of ensuring consistent equivalent width measurements. Continuum placement is a major source of uncertainty for any equivalent width measurement (Stetson & Pancino 2008). Although it is time-consuming, the best method to determine the local continuum is done manually using the interactive mode. Our "autocfeature" program performs a cubic spline interpolation between user-selected anchor points to fit the local continuum used for the measurement. Pre-specified wavelength values (Table 4) are then matched to the closest wavelength values in the "pseudo-continuum," and the program calculates the equivalent width. The ability to consistently pick wavelength values is especially important when measuring strongly blended lines (Figure 1).

Zoom In Zoom Out Reset image size

Since the projected rotational velocity (v sin i) of the confirmed Lambda Boo stars range from 3 to 240 km s⁻¹, we also performed tests on the effect of stellar rotation on the line profile and our equivalent width measurements. Although the larger v sin i does "broaden" the line profile, the equivalent width measurement remains the same (less than 1% difference) as long as the wavelength range we specified for measuring the equivalent width (Table 4) covers all of the "broadened" absorption lines.

SPECTRUM is a stellar spectral synthesis program that requires the following inputs: a user generated model atmosphere (Section 3.2), an atomic/molecular line list (Section 3.3), an abundance table specifying abundance values for each element and molecule, and a value of microturbulent velocity (v_turb). SPECTRUM carries out its computation under the assumptions of LTE and a plane parallel atmosphere.

In order to create a new model atmosphere with the desired stellar parameters, we ran the newest version of ATLAS9 (revised 2010 November 8th).¹¹ Inputs required by ATLAS9 include an opacity distribution function (ODF) file (Castelli & Kurucz 2004), a Rosseland opacity file, and an initial model atmosphere to use as a baseline (Kurucz 1970). More information about ATLAS9 and our modeling procedure can be found in Section 4.1.

In order to reproduce a high-resolution spectrum, a reliable list of absorption lines in the spectral range of interest and an accurate set of atomic and molecular parameters are needed. Since line lists for wavelengths <2000 Å are not available for download with SPECTRUM, we had to compile a new line list for the wavelength range of 1000–2000 Å. We created this list using the Kurucz line list (Kurucz et al. 1995) and the National Institute of Science and Technology (Kramida et al. 2014) atomic line database. We included all available lines from these sources in our wavelength range of interest between 1000 and 2000 Å. This UV line list contains approximately 930,000 atomic and molecular lines.

Heiter (2002) used abundances of 34 Lambda Boo stars to construct a "mean" Lambda Boo abundance pattern. She concluded that the iron-peak elements from Sc to Fe as well as Mg, Si, Ca, Sr, and Ba are depleted by about 1.00 dex relative to the solar chemical composition. Al is ∼0.50 dex more depleted than Fe. The abundances of the light elements C, N, O and S are around the solar values.

As Lambda Boo stars have been characterized by strong depressions near 1600 Å (see Section 4.3), it is critical to accurately represent that region when creating synthetic spectra for Lambda Boo stars. The most recent ATLAS9 ODFs computed by Castelli & Kurucz (2004) account for line blanketing of the Lyα H–H and H–H⁺ quasi-molecular absorptions near 1600 Å (Castelli & Kurucz 2001). SPECTRUM also performs calculations to account for those absorptions.

Using ATLAS9 and SPECTRUM, we first conducted a detailed "proof of concept" study, reproducing the observed spectra of two test cases with well-determined parameters (see Section 4.1). We then generated six sets of models that are representative of:

• 1.  
  "Mean" Lambda Boo stars with [M/H] = −1.00, log g = 4.00, and elemental abundances as provided in Table 8 of Heiter (2002).
• 2.  
  "Normal" stars with [M/H] = 0.00, log g = 4.00, and solar abundances from Grevesse & Sauval (1998).
• 3.  
  "Metal-poor" stars with [M/H] = −0.50, log g = 4.00.
• 4.  
  "Metal-poor" stars with [M/H] = −1.00, log g = 4.00.
• 5.  
  "Metal-poor" stars with [M/H] = −1.50, log g = 4.00.
• 6.  
  "Metal-poor" stars with [M/H] = −2.00, log g = 4.00.

For model sets three through six the elemental abundances are reduced by the given [M/H] values with respect to solar abundances. All six sets of models were generated using a microturbulence velocity of 2.0 km s⁻¹. Although microturbulence velocity is a function of T_eff, based on Gebran et al. (2014) and our own analysis of how microturbulence velocity affects model spectra, we decided that a value of 2.0 km s⁻¹ is a good representation of the expected microtubulence velocity values for the range of effective temperatures in our model sets.

For each set of models we generated 20 synthetic spectra with an effective temperature range of 6000–11,000 K in a step size of 250 K. These models cover the wavelength range 1000–10000 Å. For each model and observed spectrum, several members of our team independently measured the equivalent widths of selected lines used in this study (following the procedure specified in Section 2.5). From these measurements we have determined internal error bars in order to quantify measurement repeatability. The equivalent width measurements of the 120 model spectra were internally consistent to better than 2% for all selected features.

In order to test the validity of our line lists, atmospheric models, and spectrum synthesis procedures, we first carried out a detailed "proof of concept" study motivated, in part, by Fitzpatrick's study (Fitzpatrick 2010) of the ultraviolet spectrum of Vega (a "suggested" mild Lambda Boo star; see Section 4.5). We were able to reproduce Fitzpatrick's results and also apply our modeling procedures to Lambda Boötis.

Modeling the complex UV line spectra (as revealed by the IUE high-dispersion data) of Lambda Boo (Figure 3) and other Lambda Boo stars has allowed us to confirm the specific surface abundances for C, N, O, S, and the iron peak elements. We fit the IUE spectra of Lambda Boötis and Vega. For each star, we computed a rotationally and instrumentally broadened LTE synthetic UV spectrum using a stellar atmosphere model we generated with ATLAS9 as the input into SPECTRUM. All synthetic spectra were convolved to the resolution of the IUE high-dispersion spectra. We then compared them with each star's observed IUE spectra (co-added when possible to increase signal-to-noise). Figure 2 outlines the procedure we used to derive a best-fit synthetic spectrum. A step-by-step summary of this procedure is as follows:

• 1.  
  We searched the literature for stellar parameters required to run ATLAS9 and SPECTRUM : T_eff, log g, [M/H], microturbulence velocity (v_turb), elemental abundances, projected rotational velocity (v sin i), and a limb darkening coefficient (L.D.C.). The L.D.C. used for all models was 0.6, per the general recommendation in the SPECTRUM manual.¹²
• 2.  
  ATLAS9 needs a starting model (baseline) to calculate the new model atmosphere. To determine which model to use as the starting model, we compared the observed spectrum to several model fluxes from Castelli's website with T_eff, log g, [M/H] and v_turb close to the values found in the literature.¹³ We then chose the model flux that best agreed with the observed general flux distribution and used its corresponding model atmosphere file as our starting model.
• 3.  
  We ran ATLAS9 using the starting model found in step two to create a new model atmosphere with the T_eff and log g values found in the literature.
• 4.  
  We ran SPECTRUM using the new model atmosphere generated in step three to create a synthetic spectrum for comparison with the observed spectrum. We also used ${SPECTRUM}$'s auxiliary programs AVSINI and SMOOTH2 to rotationally broaden the synthetic spectrum and convolve it to the observed spectrum's resolution.
• 5.  
  If the general flux distribution of the synthetic spectrum matched well with the observed spectrum, we proceeded to step six to refine the [M/H] value. If it did not match well, we fine-tuned the T_eff and log g values, and repeated steps three through five.
• 6.  
  In order to fine-tune the overall metallicity ([M/H]) and v_turb of the desired model atmosphere, we generated new ODFs and Rosseland opacity files (ROFs) by interpolating those files published by Castelli and Kurucz (using their supplemental programs DFINTERP and KAPROSSINTERP).
• 7.  
  We ran ATLAS9 using the new ODFs and ROFs from step six, as well as the T_eff, [M/H] and log g values from step five.
• 8.  
  We ran SPECTRUM using the model atmosphere from step seven to generate several spectra with various published elemental abundances and v sin i values.
• 9.  
  We compared the synthetic spectra generated in step eight to the observed spectrum, and then determined which synthetic spectrum fits best.

Zoom In Zoom Out Reset image size

We have demonstrated that the synthetic spectra we produced, using published elemental abundances and physical parameters (Table 5), provide an accurate representation of the observed spectra of Vega and Lambda Boo. Since Lambda Boo's observed UV spectra can be fit well (both in detail and in its broad properties) by a single-temperature synthetic spectra, models can be used to study the UV characteristics of Lambda Boo stars as a function of effective temperature (Sections 4.2 and 4.3).

Table 5.  Stellar Parameters of Lambda Boo and Vega

Property	Lambda Boo	Vega
	(HD 125162)	(HD 172167)
Teff	8700	9550
log g	4.2	4.0
vturb (km s−1)	2.0 (7)	2.04 (1)
v sin i (km s−1)	123 (4)	21.8 (8)
vrad (km s−1)	−7.9 (3)	−20.6 (3)
[M/H]	−2.00	−0.50
	Elemental Abundancesa		Solar Abundances (2)
C	−3.85 (5)	−3.86 (1)	−3.49
N	−3.93 (5)	−4.13 (1)	−4.07
O	−3.56 (5)	−3.56 (1)	−3.21
Al	−8.07b	−6.47 (1)	−5.57
S	−5.01 (6)	−5.43 (1)	−4.71
Fe	−6.42 (5)	−5.26 (1)	−4.54

Notes.

^aAll elemental abundances given as $\mathrm{log}(A/{N}_{\mathrm{total}})$. ^bThere is no published Al abundance for Lambda Boo. Our Al abundance value was scaled down from the solar Al abundance by Lambda Boo's [M/H] value (−2.0 dex), plus an additional 0.5 dex (Al is  0.5 dex more depleted than Fe in Lambda Boo stars (Heiter 2002)).

References. (1) Fitzpatrick (2010), (2) Grevesse & Sauval (1998), (3) Gontcharov (2006), (4) Royer et al. (2002), (5) Venn & Lambert (1990), (6) Heiter et al. (2002), (7) Castelli & Kurucz (2001), (8) Gulliver et al. (1994).

Download table as:  ASCIITypeset image

The main differences between Lambda Boo stars and other metal-poor stars are (1) although otherwise metal weak, Lambda Boo stars' C, N, O, and S abundances are near solar, and (2) Al is ∼0.50 dex more depleted than Fe in Lambda Boo stars (but not in stars that are just metal weak (Heiter 2002)). This abnormal abundance pattern makes the C i/Al ii equivalent width ratio a powerful tool for distinguishing Lambda Boo stars from both normal stars and other metal-poor stars. In this study we chose to measure the C i 1657 Å/Al ii 1671 Å equivalent width ratio based on the availability and quality of UV data.

We have opted not to discuss the other two equivalent width ratios used by Solano & Paunzen (1999), namely C i 1657 Å /Si ii 1527 Å and C i 1657 Å/Ca ii 1839 Å, because the Si ii and Ca ii lines are much weaker than Al ii. This results in significantly more measurement error, and based on our measurements those ratios are not as robust as the C i 1657 Å/Al ii 1671 Å equivalent width ratio.

We have examined all of the existing IUE data for our program Lambda Boo stars and found that their low-dispersion data cannot be used to resolve and measure the much weaker Al ii 1671 Å line. We therefore only include nine of our program Lambda Boo stars with IUE high-dispersion spectra in Figure 4.

Zoom In Zoom Out Reset image size

Although both C i 1657 Å and C i 1931 Å lines were observed with IUE's SWP camera (1145–1975 Å), the data quality longward of 1900 Å is generally very poor. Since both the C i 1657 Å and Al ii 1671 Å features are blended lines (Table 4), it is very important to make sure that the equivalent width measurements are inclusive of all of the necessary lines, regardless of any rotational broadening effects, and that this is repeatable for all measurements (see Section 2.5). For a typical Lambda Boo star, the C i 1657 Å blend is very strong and the Al ii 1671 Å blend is relatively weak, so the accuracy of the C i 1657 Å/Al ii 1671 Å line ratio heavily depends on the accuracy of the equivalent width measurement of Al ii 1671 Å.

We also measured the C i 1657 Å/Al ii 1671 Å equivalent width ratios of a group of normal (standard) stars with different effective temperatures. Since late A-type stars and early F-type stars do not have much flux shortward of 1700 Å, it is hard to find standard (normal) stars cooler than A7 V (∼7920 K) with existing IUE SWP high data. To help alleviate this problem, we generated a set of synthetic models for normal stars ([M/H] = 0.00) with effective temperatures ranging from T_eff = 6000 to 11,000 K in steps of 250 K. Figure 4 clearly shows a good general agreement between the observed normal stars and the synthetic models for normal stars (Table 2). It also illustrates how accurately the equivalent width ratio must be determined in order to discriminate between Lambda Boo stars and other metal-poor stars. For example, around 9500 K the error bars can be large, whereas below 8000 K a large error bar could lead to a mis-classification if the measured ratio is close to the Heiter mean (solid line in Figure 4).

In the past, some stars have been (mis)classified as Lambda Boo-type based on their metal weakness alone. In order to show that the C i/Al ii ratio is capable of separating Lambda Boo stars from other metal-poor stars, we generated a set of synthetic models for "normal" metal-poor stars. Between 7500 and 10,000 K, the C i/Al ii equivalent width ratios of Lambda Boo stars are very different from the C i/Al ii ratios of metal-poor models. We also compared the C i/Al ii equivalent width ratios of Lambda Boo stars with Heiter's "mean" Lambda Boo star models discussed in Section 3.4. Our nine Lambda Boo stars do not match with the "mean" Lambda Boo star models (with [M/H] = −1.00) exactly, because they have [M/H] ranging from −0.58 to −2.00 (Table 1). The stellar parameters of individual stars (log g, v_turb, etc.) also vary slightly from our model values of log g = 4.00 and v_turb = 2.0 km s⁻¹. Figure 4 clearly shows the Lambda Boo stars' C i 1657 Å/Al ii 1671 Å equivalent width ratio is a function of both the star's effective temperature and metallicity. One cannot use a single C i/Al ii cut-off value as a Lambda Boo classification criterion, but with the models we generated with various effective temperatures and metallicities, we can specify the C i/Al ii cut-off value for a specific star based on its effective temperature and metallicity. For all effective temperatures in the range of 7300–10,000 K, a C i/Al ii ratio >3 is a firm indication that the star is a Lambda Boo. For stars with a ratio between 2 and 3, it is important to first determine the metallicity and effective temperature to see if the ratio falls above the expected value. Stars with solar abundances all have ratios $\leqslant 2$. For example, there are two Lambda Boo stars in Figure 4 with ratios well below the "Heiter's mean" (which is based on [M/H] = −1.0; solid line) but nevertheless well above the value expected for their lower metallicities.

Baschek et al. (1984) use low and high resolution IUE-spectra to distinguish Lambda Boo stars from normal stars. The strong depression at 1600 Å (Figure 5) is one of the main criteria they used to identify Lambda Boötis stars. However, the 1600 Å depression has also been detected in field horizontal branch stars and cool DA white dwarfs (Jaschek et al. 1985).

Zoom In Zoom Out Reset image size

Our goal is not to determine the cause of this so-called "1600 Å" depression (Holweger et al. 1994), but rather to assess its use as a Lambda Boo classification criterion. Using model spectra we show in Figure 6 that all metal-poor stars with effective temperature of 8000–9000 K and log g near 4.0 also have a detectable 1600 Å depression. Therefore, a detectable 1600 Å depression is not a unique Lambda Boo classification criterion.

Zoom In Zoom Out Reset image size

Figures 5 and 6 show that the UV excess continuum in metal-poor and Lambda Boo spectra shortward of 1550 Å makes the apparent depression centered at 1600 Å more clearly visible. This blanketing effect is highly dependent on the metallicity; the lower the [M/H], the higher the general flux level in the ultraviolet, but it is the other way around in the visible/IR region (Figure 7). Shortward of 1550 Å, there is a significant difference between metal-poor and Lambda Boo stars. The latter have higher C, N, O, and S abundances, so in this spectral region the Lambda Boo stars are fainter due to the higher C i continuous opacity and higher carbon line opacity. This provides another constraint, in addition to the C i/Al ii equivalent width ratio, to discriminate between Lambda Boo stars and other metal-poor stars in the ultraviolet.

Zoom In Zoom Out Reset image size

We inspected IUE low-dispersion spectra of 27 Lambda Boo stars and found that only 8 of them show the 1600 Å depression (Table 6). We found that not all Lambda Boo stars have detectable 1600 Å depressions and that the presence of a depression is not enough to discern a Lambda Boo star from other metal-poor stars.

Ohanesyan (2008) demonstrates that peculiar A-type stars—including Lambda Boo stars—can be identified by analyzing the equivalent width of the 2786–2810 Å Mg ii h+k spectral band as a function of the star's T_eff. We measured the equivalent width of the 2794.1–2805.3 Å spectral band of 20 Lambda Boo stars with suitable IUE spectra, as well as all of our model spectra. We chose a slightly narrower band than Ohanesyan (2008) because we only want to measure the total equivalent width of Mg ii k (2796.4 Å in vacuum) and h (2803.5 Å in vacuum) resonance lines (Table 4). The observed spectra have a small amount of additional absorption due to interstellar, and possibly circumstellar material, which is not included in the synthetic spectra (Figure 8). For rapidly rotating stars with high resolution spectra, it is possible to remove the contribution to the overall equivalent width from this narrow component. Because it is not possible to do this in all the spectra, and because its contribution to the overall equivalent width in this 11 Å bandpass is small, we did not attempt to make any corrections to the values shown in Figure 9.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For those Lambda Boo stars with existing ELODIE data, we also measured the equivalent width of Mg ii 4481 Å (Figure 8) and compared them with three sets of models (the [M/H] = −2.00 models, the [M/H] = −1.00 models, and the solar [M/H] = 0.00 models). All the synthetic spectra were convolved to the resolution of the ELODIE spectra (R = 42,000). Figure 9 clearly shows that the equivalent width of Mg ii 4481 Å is accurately predicted by our models, and it is a function of both the star's T_eff and [M/H], and one cannot use the weak Mg ii 4481 Å line alone to distinguish Lambda Boo stars from other metal-poor stars. Furthermore, the temperature dependence of the Mg ii 4481 Å equivalent width mimics that of the 2800 Å Mg ii feature. Another widely accepted visible characteristic of Lambda Boo stars is that "the ratio Mg ii 4481 Å/Fe i 4383 Å is significantly smaller than in normal A-type stars" (Gray & Corbally 2009). However, the weaker Mg ii 4481 Å/Fe i 4383 Å equivalent width ratio was judged by eye using classification-resolution spectra, which could include blends with many other lines. We have measured the equivalent widths of the blended 4383 Å feature in both our grid of models and the higher-resolution ELODIE spectra. The Fe i 4383 Å line is very weak, especially for hotter stars. Even in the noise-free models it is hard to measure the ratio for effective temperatures greater than 9500 K. For stars with v sin i >80 km s⁻¹, rotational broadening makes it more difficult to measure the lines' equivalent width, so the measurement error is large. We have, however, demonstrated that a weak 4481 Å line is a good indicator that a star is metal-poor.

The Lambda Boo class has recently regained the spotlight because of the discovery of giant planets around the Lambda Boo star HR 8799 (Gray & Kaye 1999; Marois et al. 2008) and around a "suggested" Lambda Boo star Beta Pictoris (Bonnefoy et al. 2011). The discovery of a giant asteroid belt around Vega, another suggested Lambda Boo star, also raises the possibility that it could have hidden planets (Su et al. 2013). Thus, there could be a link between Lambda Boo stars and planet-bearing stars (Jura 2015). We would like to investigate whether HR 8799, Vega, and Beta Pictoris satisfy the UV criteria we established in this study. Unfortunately, there are no existing UV data for HR 8799, but we have applied the UV criteria established in this paper to Beta Pictoris and Vega.

Beta Pictoris is the original and best-known example of a star surrounded by a dusty debris disk, and it has ample existing IUE data. King & Patten (1992) claim that "Beta Pic seems to meet most criteria required for a Lambda Boo classification." On the other hand, Holweger & Rentzsch-Holm (1995) conclude that "Beta Pic most probably is a superficially normal A star with solar metallicity." Gray et al. (2006) observed Beta Pic using CTIO 1.5 m telescope with the Cassegrain spectrograph and the Loral 1200 X 800 CCD (with a 2 pixel resolution of 2.6 Å in the spectral range of 3800–5150 Å). They conclude that Beta Pictoris has a spectral type of A6 V with no peculiarities detected.

We examined all of Beta Pic's existing archival IUE data. We conclude that Beta Pic is not a Lambda Boo star because:

• 1.  
  its C i 1657 Å/Al ii 1671 Å equivalent width ratio (Figure 10) is too low for a Lambda Boo star with Beta Pic's T_eff and [M/H] (Table 7);
• 2.  
  the 1600 Å depression of a metal-deficient A-type star with T_eff = 8182 K and Beta Pic's stellar parameters should be clearly visible, but we did not detect one;
• 3.  
  its Mg ii equivalent widths are the same as normal A-type stars with Beta Pic's T_eff (Figure 10);
• 4.  
  Beta Pic's abundance pattern is like those of normal stars with solar abundances (Holweger & Rentzsch-Holm 1995).

Zoom In Zoom Out Reset image size

Vega was first noted to have Lambda Boo-like abundances by Baschek & Slettebak (1988). Since then there has been a long debate on whether Vega should be considered as a member of the Lambda Boo class. Ilijic et al. (1998) performed an elemental abundance analysis using a high-dispersion spectrum in the visible region, and they suggest that Vega is a "mild" Lambda Boo star based on its mild metal under-abundance. Although Vega does have an abundance pattern like those of mild Lambda Boo stars, with C and O near solar and metals being about 0.5 dex below solar (see Ilijic et al. 1998; Figure 7 of Hekker et al. 2009; Fitzpatrick 2010), it does not have any of Lambda Boo's UV characteristics. We conclude that Vega is not a Lambda Boo star because:

• 1.  
  it has a C i 1657 Å/Al ii 1671 Å equivalent width ratio that is too low for a Lambda Boo star with the same effective temperature as Vega (Figure 10);
• 2.  
  it does not have a 1600 Å depression though for a star with T_eff = 9562 K one should be clearly visible;
• 3.  
  both its C i 1657 Å/Al ii 1671 Å equivalent width ratio and its Mg ii 2800 Å equivalent width are similar to normal A-type stars with the same T_eff and [M/H] (−0.50 dex) (Figure 10);
• 4.  
  its Mg ii 4481 Å equivalent width is slightly above our predicted value for normal A type stars with the same T_eff and [M/H]. It should be noted that the Mg ii 4481 Å equivalent width is not a conclusive diagnostic for the Lambda Boo class. We are currently working on a visible classification criteria that can differentiate Lambda Boo stars from other metal-poor stars in a manner similar to what we have done in this paper.