Searching for interstellar C60+ using a new method for high signal-to-noise HST/STIS spectroscopy

TL;DR

This study introduces a novel high signal-to-noise spectroscopy method using HST/STIS to detect interstellar C60+ molecules, achieving unprecedented data quality and revealing potential correlations with interstellar radiation, though conclusive detection remains challenging.

Contribution

The paper presents a new ultra-high S/N spectroscopic technique with HST/STIS in scan mode, enabling better detection of interstellar molecules like C60+ without telluric contamination.

Findings

Achieved S/N of 600-800 in near-IR spectra of background stars.

Detected the 9577 Å DIB attributed to C60+ in a heavily-reddened star.

Identified a possible correlation between DIB strength and interstellar radiation field.

Abstract

Due to recent advances in laboratory spectroscopy, the first optical detection of a very large molecule has been claimed in the diffuse interstellar medium (ISM): C60+ (ionized Buckminsterfullerene). Confirming the presence of this molecule would have significant implications regarding the carbon budget and chemical complexity of the ISM. Here we present results from a new method for ultra-high signal-to-noise (S/N) spectroscopy of background stars in the near infrared (at wavelengths 0.9-1 micron), using the Hubble Space Telescope Imaging Spectrograph (STIS) in a previously untested `STIS scan' mode. The use of HST provides the crucial benefit of eliminating the need for error-prone telluric correction methods in the part of the spectrum where the C60+ bands lie, and terrestrial water vapor contamination is severe. Our STIS spectrum of the heavily-reddened B0 star BD63\,1964 reaches an unprecedented S/N for this instrument ($\sim600-800$), allowing the detection of the diffuse interstellar band (DIB) at 9577 \AA\ attributed to C60+ as well as new DIBs in the near-IR. Unfortunately, the presence of overlapping stellar lines, and the unexpected weakness of the C60+ bands in this sightline, prevents conclusive detection of the weaker C60+ bands. A probable correlation between the 9577 \AA\ DIB strength and interstellar radiation field is identified, which suggests that more strongly-irradiated interstellar sightlines will provide the optimal targets for future C60+ searches.

Full text

1. Introduction

The diffuse interstellar band (DIB) problem is the longest-standing puzzle in interstellar chemistry, seemingly impenetrable despite the dedicated efforts of astronomers and laboratory chemists since the early 20th century (Herbig, 1995; Sarre, 2006; Snow, 2014). The DIBs manifest as broad spectroscopic absorption features in the optical to near-infrared (NIR) spectra of stars as their light passes through the diffuse interstellar medium, indicating the presence of a large quantity of (mostly carbonaceous), unidentified molecular material (e.g. Cordiner, 2011). Foing & Ehrenfreund (1994) discovered two DIBs at 9577 Å and 9632 Å that they assigned to C${}_{60}^{+}$ (ionized Buckminsterfullerene) based on a similarity with the absorption wavelengths seen in neon matrix spectroscopy (d’Hendecourt et al., 1992; Fulara et al., 1993). Recently obtained gas-phase spectra of C{}_{60}^{+}$$-Hen complexes (for $n\leq 4$; Campbell et al. 2015, 2016a; Kuhn et al. 2016), show that the match between the interstellar and laboratory wavelengths is accurate to within the observational uncertainties (a fraction of an angström). The bare C${}_{60}^{+}$ absorption wavelengths were confirmed in an independent laboratory study by Spieler et al. (2017, submitted) using C${}_{60}^{+}$ cations embedded in He-droplets.

If confirmed, the detection of interstellar C${}_{60}^{+}$ will constitute a major breakthrough in interstellar chemistry and may provide, for the first time, an insight into the true scale of chemical complexity in the diffuse ISM. Measurement of the NIR electronic transitions of C${}_{60}^{+}$ (in absorption) would also provide a unique complement to the discovery of mid-infrared C60 and C${}_{60}^{+}$ emission bands in circumstellar and interstellar environments (Cami et al., 2010; Sellgren et al., 2010; Berné et al., 2013).

Despite dedicated observational studies (Walker et al., 2015, 2016; Galazutdinov et al., 2017), the case for interstellar C${}_{60}^{+}$ has not yet been proven beyond reasonable doubt. Based on the laboratory measurements, five absorption features are expected (at 9348.4, 9365.2, 9427.8, 9577.0 and 9632.1 Å, with strength ratios 0.07:0.2:0.3:1.0:0.8; Campbell et al. 2016b). Galazutdinov et al. (2017) were unable to confirm the presence of the weakest three features in a sample of 19 heavily-reddened Galactic sightlines observed from the ground. Instead of a constant ratio for the two strongest C${}_{60}^{+}$ DIBs ($\lambda$9577 and $\lambda$9632), as expected for electronic transitions arising from a ${}^{2}A_{1u}$ ground vibronic state, Galazutdinov et al. (2017) found that the interstellar band ratio was highly variable among different lines of sight. We note, however, that if the observed transitions involve lower levels above the ground state (such as the split levels arising from Jahn-Teller distortion), a variable ratio could potentially occur. Although Walker et al. (2016) claimed interstellar detections of all five C${}_{60}^{+}$ bands, the three weaker C${}_{60}^{+}$ bands are problematic as they fall in a wavelength region heavily obscured by absorption due to water vapor in the Earth’s atmosphere (see Galazutdinov et al., 2000, 2017). Telluric correction methods for such weak interstellar absorption features are error-prone as a result of incomplete telluric line cancellation, and from the possible presence of weak (unseen) stellar lines in the telluric standard spectrum. To rigorously confirm the identification of interstellar C${}_{60}^{+}$, high signal-to-noise observations of all five absorption bands are required, preferably from outside Earth’s atmosphere.

In this article, we present the first ever ultra-high S/N, high resolution stellar/interstellar spectra obtained using the Hubble Space Telescope (HST) imaging spectrograph, targeting four of the five C${}_{60}^{+}$ features ($\lambda$9349, $\lambda$9365, $\lambda$9428 and $\lambda$9577) along a heavily reddened line of sight. Very high continuum signal-to-noise ratios (S/N$>500$ per spectral channel) are commonly achieved using ground-based telescopes, but such sensitivity has not previously been obtained using (direct) HST spectroscopy due to the severe CCD fringing at red and near-IR wavelengths. After successfully employing a previously untested ‘STIS scan’ observing mode to trail the target star along the spectrograph slit (crossing hundreds of CCD rows to facilitate fringe cancellation), unprecedented signal-to-noise ratios $\sim 600$-800 have been obtained. These spectra have permitted the first search for weak interstellar absorption features in the NIR, unhindered by telluric absorption.

2. Observations and data reduction

To reliably detect the weaker C${}_{60}^{+}$ DIBs, a high-S/N interstellar spectrum is required, with minimal contamination from interloping stellar photospheric lines. The heavily-reddened ($E_{B-V}$ = 1.01) B0 I star BD+63 1964 was selected due to its extremely strong DIBs — the majority of DIBs are among the strongest for stars of comparable extinction (Tuairisg et al., 2000) — as well as its early spectral type and relatively clean predicted spectrum in the vicinity of the $\lambda$9365 and $\lambda$9428 C${}_{60}^{+}$ bands. The lightly-reddened B0 I star 69 Cygni ($E_{B-V}$ = 0.14) was selected for comparison, to assist in the discrimination of stellar and interstellar features.

2.1. HST STIS

Data were acquired over a single orbit for 69 Cyg and two orbits for BD+63 1964, using the G750M grating with a central wavelength of 9336 Å (covering the range 9050-9610 Å). This covers four of the five known C${}_{60}^{+}$ bands; a change of grating tilt would be required to obtain the fifth band. We used the $52^{\prime\prime}\times 0.1^{\prime\prime}$ slit, and the plate scale was $0.05^{\prime\prime}$ per pixel. Following peakup and focusing maneuvers, a series of STIS-scan exposures was performed for each star, trailing the star along the slit at a constant rate with the shutter held open. For 69 Cyg, six exposures were obtained: two long scans spanning up to 1000 detector rows and four shorter scans spanning $\sim 300$ rows. The scan direction was alternated for successive exposures. For BD+63 1964 (observed in a separate visit after the optimal observing sequence had been established), eight identical forward scans were obtained spanning 700 detector rows each, towards the upper part of the CCD. A pair of flat-field exposures (contemporaneous fringe flats) was obtained in sequence following each pair of target star exposures, and 5-6 additional pairs of flats were obtained to fill in the remaining orbital time during occultation. Exposures of a Pt/Cr-Ne lamp were obtained during each orbit for wavelength calibration. Subsequent conversion from vacuum to air wavelengths was performed using the Morton (2000) formula.

The STIS CCD suffers from charge transfer inefficiency (CTI) during readout, as a result of long-term radiation damage. The raw frames were corrected for CTI using the procedure outlined by Anderson & Bedin (2010)111see http://www.stsci.edu/hst/stis/software/analyzing/scripts/pixel_based_CTI, which significantly reduces the severity of bad CCD columns. Standard dark current, bias, geometric distortion and wavelength corrections were performed using the Iraf calstis package (Hodge et al., 1998). For BD+63 1964, the four trailed science exposures in each orbit were sufficiently well matched to allow statistical cosmic ray rejection using ocrreject. This was not possible for 69 Cyg due to imperfect registration of the exposed rows between each frame, which may have occurred due to errors in the guide star tracking, as well as possible exposure timing errors.

Spectral extraction was performed by summing the counts along CCD columns. Bad pixels and residual cosmic rays were rejected based on their standard deviation (using a $3\sigma$ threshold), after first normalizing the individual CCD rows by their median levels. Such row normalization was required due to variations in the illumination function along the slit, resulting from slight variations in the slit width and telescope scan rate. Scattered light subtraction was performed with a low-order fit to the light under the two occulting bars (positioned 1/3 and 2/3 way along the slit). Extracted Pt/Cr-Ne lamp line widths were consistent with the nominal instrumental resolving power of 10,000 (30 km $\rm{s}^{-1}$).

The STIS CCD suffers from severe fringing due to internal reflections at wavelengths greater than 700 nm, where the chip starts to become transparent to incoming light. Fringe amplitudes are variable with wavelength, with a maximum of $\pm 20\%$ near 900 nm (see e.g. Figure 1). The aim of our fringe cancellation strategy was to match as closely as possible the illumination pattern of the science exposure to that of the flat field, by STIS-scanning. This approach has the major advantages of (1) reducing the impact of individual bad pixels and cosmic ray hits, (2) allowing for orders of magnitude increases in the number of counts per science exposure, (3) eliminating the need to change the slit mechanism during observations, resulting in a consistent optical setup for the science and flat-field exposures.

As shown in Figure 1, scanning the exposed star across hundreds of CCD rows results in significantly reduced fringe amplitudes (a five-fold reduction in the spectral RMS in this case). Following correction using a contemporaneous fringe flat, we obtained apparently complete fringe cancellation down to the level of the statistical noise.

2.2. CFHT ESPaDOnS

For comparison with our STIS data, four 1000 s exposures of BD+63 1964 were obtained over 2 nights in 2016 December, using the ESPaDOnS échelle spectrograph of the Canada-France-Hawaii Telescope (CFHT). Observing conditions were good and the airmass was in the range of 1.41-1.49. The data were reduced by the automated Upena pipeline, which uses the Libre-ESpRIT data reduction software (Donati et al., 1997). The reduced spectra cover the range 370-1048 nm at a resolving power of $\sim$80,000, and have a continuum signal-to-noise ratio $\sim 500$ at 900 nm.

Due to the forest of atmospheric H2O absorption lines in our STIS range, the ESPaDOnS spectra need to be telluric-corrected. The conventional approach of division by a standard star is unreliable for the measurement of weak DIBs due to the likely presence of stellar features in the standard star spectrum, which would introduce artifacts upon division. We therefore adopted a transmission modeling approach, using synthetic atmospheric transmittances provided by the TAPAS website222http://ether.ipsl.jussieu.fr/tapas/ (Bertaux et al., 2014), adapted for Mauna Kea. A preliminary telluric correction was performed using the ‘rope length’ method for weak to moderately strong lines (Raimond et al., 2012), excluding from the length the spectral intervals corresponding to the central parts of the strongest lines. The preliminary correction was then interpolated in the most heavily contaminated intervals to remove residual artifacts. Next, the observed data were fitted to the convolved product of this spectrum and the TAPAS synthetic transmittance, allowing for variable airmass, refined wavelength correction and variable line-spread function. The observed data were divided by this new, adjusted transmission spectrum to provide a new corrected spectrum. The second step was then iterated to convergence to produce the final corrected spectrum. Despite our best efforts, telluric residuals still remain at the locations of the most heavily saturated lines, which may be due to time-variability of the telluric spectrum, scattered light residuals or small asymmetries in the ESPaDOnS instrumental response function.

2.3. Comparison between STIS and ESPaDOnS spectra

The main purpose of these ESPaDOnS spectra was to verify the reliability of our nonstandard STIS observation and reduction procedures. A comparison between the HST/STIS and telluric-corrected CFHT/ESPaDOnS spectra is shown in Figure 2. Telluric-correction artifacts in panels (a) and (b) are severe, preventing a reliable comparison in the vicinity of the weaker C${}_{60}^{+}$ bands. However, in the less problematic range 9500-9600 Å (panel c), the agreement between the two instruments is excellent, which demonstrates the success of our STIS observation and data reduction strategy. Slight discrepancies may be due to differences in the instrumental resolving power, telluric correction residuals or uncertainties in the continuum level of the high-dispersion CFHT spectrum. Some uncertainty in the STIS scattered light correction also remains, leading to uncertainty in the zero level. More detailed characterization of the (2D) scattered light properties for STIS scan exposures awaits further study.

3. Results

The combined, scattered-light-subtracted exposures of BD+63 1964 and 69 Cyg reached total counts of $8\times 10^{5}$ and $5.5\times 10^{5}$, respectively (per dispersion pixel). The presence of weak stellar and interstellar features precludes an accurate measurement of the continuum RMS, but in regions free of any obvious lines, S/N = 600-800 was found for BD+63 1964 and $\sim 700$ for 69 Cyg. To our knowledge, this is the highest S/N ever demonstrated with STIS for direct near-infrared spectroscopy at full grating resolution. The continuum-normalized spectra are shown in Figure 3.

Both stars are B0 supergiants and have nearly identical stellar spectra. We used the non-LTE model atmosphere code CMFGEN (Hillier & Miller, 1998, 1999; Hillier, 2011), which solves the radiative-transfer equation for a spherically symmetric wind in the co-moving frame under the constraints of radiative and statistical equilibrium. An effective temperature of 27,000 K, macro-turbulent velocity of 120 km $\rm{s}^{-1}$ and projected rotational velocity of 85 km $\rm{s}^{-1}$ was derived for both stars. The modeled lines are identified in Figure 3. Four features stand out in the BD+63 1964 spectrum that are not present in 69 Cyg or in the stellar model, which are likely due to interstellar absorption. The first three of these (at 9088, 9302 and 9412 Å) are identified as new diffuse interstellar bands; Galazutdinov et al. (2000) also found a possible DIB near 9412 Å. The absorption feature at 9577 Å coincides with the laboratory C${}_{60}^{+}$ wavelength. 69 Cyg also shows evidence for weak absorptions at 9577 and 9412 Å.

The equivalent width (EW) of the $\lambda$9577 band is $74\pm 3$ mÅ. A Gaussian fit to its profile gives EW = $76\pm 3$ mÅ, with a central wavelength of $9576.60\pm 0.03$ Å, a central depth of 2.3% and FWHM = $3.1\pm 0.1$ Å. These line strength measurements may be considered accurate to within about $\pm 10$% due to uncertainties in the STIS scattered light subtraction.

A close-up view of the four observed C${}_{60}^{+}$ band regions is shown in Figure 4. A spectral model for C${}_{60}^{+}$ is overlaid using the laboratory wavelengths and band strengths of Campbell et al. (2016a). We adopt Gaussian band shapes because a Lorentzian profile provides a poor match to the observed $\lambda$9577 band wings (Fig. 4). The model has been Doppler shifted to match the interstellar K i velocity. As with most heavily-reddened diffuse sightlines, the K i line towards BD+63 1964 (recorded in our ESPaDOnS data) shows complex velocity structure, with three main components at $-16$, $-27$ and $-33$ km $\rm{s}^{-1}$. We take the average of these ($-25$ km $\rm{s}^{-1}$) as the interstellar C${}_{60}^{+}$ radial velocity. The $\lambda$9577 band carrier abundance could be variable across the three K i clouds (due to variations in chemical abundances and ionization levels), so we assign an error margin of $\pm 10$ km $\rm{s}^{-1}$, which corresponds to $\pm 0.3$ Å. Within the uncertainties, our inferred (Doppler corrected) $\lambda$9577 rest wavelength of $9577.4$ Å matches the laboratory value of $9577.0\pm 0.2$ Å; the slight redshift could indicate that C${}_{60}^{+}$ is more associated with the higher-velocity K i gas. The 1.0 Å instrumental resolution contributes 0.2 Å to the observed DIB FWHM. The additional (0.4 Å) of broadening in the observed band compared with the laboratory measurement of 2.5 Å is consistent with the spread of interstellar K i velocities in this sightline.

We find weak absorption features in our BD+63 1964 spectrum close to the wavelengths of the $\lambda$9348, $\lambda$9365 and $\lambda$9428 C${}_{60}^{+}$ bands. These features are not present towards our standard star 69 Cyg, and may therefore be due to interstellar C${}_{60}^{+}$. However, the appearance of the $\lambda$9348 band is far from clear due to contamination by another possible weak DIB at 9350 Å (previously identified at a rest wavelenth of 9351 Å by Walker et al. 2016), as well as overlapping stellar He ii absorption. The $\lambda$9365 feature is partially contaminated by a weak stellar O ii line, and is broader and deeper than expected. The $\lambda$9428 feature is too weak for a reliable measurement.

4. Discussion

The strengths of different diffuse interstellar bands are known to vary among sightlines with differing physical and chemical properties, independent of the total amount of interstellar material (e.g. Krełowski & Walker, 1987; Ehrenfreund & Jenniskens, 1995; Cami et al., 1997; Ensor et al., 2017; Elyajouri et al., 2017). The equivalent width per unit reddening (EW/$E_{B-V}$) provides a measure of the relative strength of a given DIB. For BD+63 1964  we find EW(9577)/$E_{B-V}$ = 73 mÅ, which is among the smallest values known for this band. Unfortunately for our present study, this means that BD+63 1964 is one of the least favorable sightlines in which to search for C${}_{60}^{+}$, which explains to some degree our difficulties in establishing (or disproving) the presence of the weaker C${}_{60}^{+}$ absorption bands.

The extreme weakness of the $\lambda$9577 band may be due to a lower than normal degree of ionization in this sightline. The well-known DIB at 6283 Å has a large EW/$E_{B-V}$ value in diffuse clouds that are strongly irradiated by UV (see above references), while the 5797 Å DIB tends to be favored in more neutral, less irradiated clouds. The $\lambda$5797/$\lambda$6283 equivalent width ratio is thus considered to be a tracer of the radiation field strength in the line of sight. Data have been taken from compilations of optical DIB measurements by Xiang et al. (2017) and the $\lambda$9577 measurements of Galazutdinov et al. (2000) to study the relationship between EW(9577)/$E_{B-V}$ and the $\lambda$5797/$\lambda$6283 DIB strength ratio. Additional $\lambda$6283 and $\lambda$5797 measurements were obtained from archival data from the CFHT and Telescope Bernard Lyot. The results for 9 sightlines are shown in Figure 5, including the present results for BD+63 1964. Even without HD 37022 (a highly-irradiated sightline through the Orion Nebula with unusually high EW(9577)/$E_{B-V}$), the trend is for increasing $\lambda$9577 strength in gas pervaded by stronger radiation fields. In a survey of DIBs and atomic lines towards BD+63 1964 by Ehrenfreund et al. (1997), a relatively high gas density and low level of ionization was inferred, and the extremely low $\lambda$5797/$\lambda$6283 ratio is suggestive of weakly-irradiated gas. The trend for increasing EW(9577)/$E_{B-V}$ with radiation field strength is consistent with the assignment to C${}_{60}^{+}$ or another species whose ionisation stage becomes dominant in more irradiated diffuse clouds (see also Foing & Ehrenfreund, 1997).

5. Conclusion

We performed the first successful test of the HST ‘STIS scan’ spectroscopic observational mode. The resulting spectroscopic S/N (up to $\sim 800$) is, to our knowledge, by far the highest demonstrated with STIS for direct stellar measurements at full grating resolution. This technical demonstration opens up a new wavelength range for interstellar, stellar and exoplanetary spectroscopy, with unprecedented sensitivity.

We obtained the first ‘clean’ measurements of the spectral region covering the $\lambda$9348, $\lambda$9365, $\lambda$9428 and $\lambda$9577 C${}_{60}^{+}$ bands, without the confounding presence of telluric contamination or fringing artifacts. The $\lambda$9577 band exhibits a closely Gaussian shape with FWHM and peak wavelength consistent with laboratory measurements. However, due to the surprising weakness of the C${}_{60}^{+}$ features in our chosen sightline, the strengths and profiles of the (intrinsically weaker) $\lambda$9348, $\lambda$9365 or $\lambda$9428 bands could not be reliably measured. We identify a correlation between the $\lambda$9577 band strength (per unit reddening) and the $\lambda$5797/$\lambda$6283 DIB strength ratio (an indicator for radiation field strength), which suggests that the weakness of $\lambda$9577 towards BD+63 1964 is due to a low-radiation environment, consistent with the assignment of this DIB to an ion of a species whose first ionization energy is much less than that of neutral hydrogen (13.6 eV), and whose second ionization energy is $\gtrsim 13.6$ eV. Using the STIS scanning technique, high-S/N HST spectroscopy of more strongly UV-irradiated, heavily reddened sightlines may provide the best opportunity for a conclusive identification of all five C${}_{60}^{+}$ bands, which will be required to place the detection of this molecule beyond reasonable doubt.

Possible new DIBs are reported at 9088, 9302 and 9412 Å. Their interstellar origin may be confirmed by followup studies of stars with differing spectral types and degrees of extinction.

Based on observations made with the NASA/ESA Hubble Space Telescope (program #14705). Support was provided by NASA through a grant from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-26555. F.N. acknowledges Spanish grants FIS2012-39162-C06-01, ESP2013-47809-C3-1-R and ESP2015-65597-C4-1-R. ESPaDOnS observations were obtained under program 16BD86 at the Canada-France-Hawaii Telescope, which is operated by the National Research Council of Canada, the Institut National des Sciences de l’Univers of the Centre National de la Recherche Scientifique of France, and the University of Hawaii.