AN ANALYSIS OF THE ENVIRONMENTS OF FU ORIONIS OBJECTS WITH HERSCHEL^*

The basic characteristics of the FUors in our sample are listed in Table 2. Although all are classified as FUors, our sources were selected to span a variety of subtypes, with regard to local extinction, dust properties, observed eruption date, and SED shape (Green et al. 2006; Quanz et al. 2007c; Kóspál et al. 2011). Pre-Herschel observations of HBC 722 and FU Ori suggest that they fall toward the disk-like end of the spectrum (based on their mid-IR colors and low extinction values), with low (∼0.01–0.02 M_☉) upper limits on envelope mass (Sandell & Weintraub 2001; Dunham et al. 2012). V1735 Cyg is most similar to an embedded source with a highly extinguished central star and a rising mid-IR SED. V1331 Cyg, V1057 Cyg, and V1515 Cyg are intermediate between the FU Ori and V1735 Cyg cases: moderate extinction values, relatively blue SEDs but clear evidence for substantial envelope material. All but V1331 Cyg were identified as FUors by their optical flare, rather than by ancillary spectral characteristics. Typically, FUors are classified based upon their optical and mid-IR spectral features; in this work, we also classify the sources by their full SED properties.

FU Ori. Located in Orion, FU Ori (600 pc) is the archetype of the FUor class, the earliest observed outburst and the most extreme, rising by 6 mag in B over a period of 3 months in 1936. It has since faded by only ∼1 mag, at a rate of 14 mmag yr⁻¹ (Kenyon et al. 2000). FU Ori shows a relatively blue SED (Hartmann & Kenyon 1996; Green et al. 2006; Quanz et al. 2007c), with emission from pristine silicate dust in the Spitzer bands, and a relatively low extinction (A_V = 1.8 mag). Models of the FU Ori system (with a central stellar mass of 0.3 M_☉) suggest that the current accretion rate remains very high (10⁻⁴ M_☉ yr⁻¹) and it is the northern component in a close (05) binary system (Reipurth & Aspin 2004; Wang et al. 2004). The non-outbursting star, FU Ori S, may be a more massive companion (> 0.5 M_☉; Pueyo et al. 2012); the binary is unresolved by Herschel.

V1735 Cyg. Also known as Elias 1-12, V1735 Cyg (950 pc) is the only source in our sample that does not show silicate emission features in its IRS spectrum, but instead shows absorption due to CO and H₂O ice. It is the most extinguished source in the sample, with estimates of A_V > 6 mag. A second submillimeter source, V1735 Cyg SM1, is nearby (∼20''–24'' distant; Sandell & Weintraub 2001). The mid-IR SED suggests substantial envelope emission in the system.

V1515 Cyg. Gradually rising during the 1940s and 1950s, V1515 Cyg (1000 pc) shows a flat mid-IR SED with weak silicate emission features and A_V = 3.0 mag. The region around V1515 Cyg shows an arc-like reflection nebula structure (Kóspál 2011).

V1057 Cyg. Erupting around 1969, V1057 Cyg (600 pc) is one of two sources in the FOOSH sample with a pre-outburst optical spectrum consistent with a T Tauri star (Herbig 1977). With A_V = 3.5 mag and a flat-spectrum SED, weak silicate emission, and a ring-like reflection nebula, V1057 Cyg is quite similar in character to V1515 Cyg in its optical/IR spectrum.

V1331 Cyg. Classified as "pre-outburst" or between outbursts (Welin 1977), V1331 Cyg (550 pc) was identified based on its spectral similarity to the pre-outburst optical spectrum of V1057 Cyg; no optical flare was observed, and thus it is possible that V1331 Cyg is not an FUor (Sandell & Weintraub 2001; Quanz et al. 2007a; Kóspál et al. 2011). V1331 Cyg is the least luminous of the Cygnus sources in our sample. Hot water vapor was observed at the K band and was attributed to the inner disk (Najita et al. 2009). We present archival IRS spectra of this source that closely resemble the flat-spectrum SEDs of V1515 Cyg and V1057 Cyg, with weak, pristine silicate emission features at 10 and 20 μm. Even if V1331 Cyg is not an FUor, it represents an important historical comparison; the accretion rate would be very high for a non-outbursting T Tauri star at ∼10⁻⁶ M_☉ yr⁻¹ (Najita et al. 2009).

HBC 722. Also known as LkHα 188-G4, PTF10qpf, and V2493 Cyg, HBC 722 (520 pc) is a newly erupted FUor (Semkov & Peneva 2010; Miller et al. 2011; Semkov et al. 2012) located in the "Gulf of Mexico" in the southern region of the North American/Pelican Nebula, among a closely packed set of accreting young stars. Pre-outburst observations (Cohen & Kuhi 1979) indicate that HBC 722 was a ∼0.5 L_☉ K7–M0 spectral-type T Tauri star prior to 2009. Between 2009 and 2010, HBC 722 rose by ∼1 mag in B, and then suddenly flared between 2010 July and September, rising by 3.5 mag and in luminosity by a factor of ∼20 to 12 L_☉ over the course of 3 months (Kóspál et al. 2011). After decreasing to 5.4 L_☉ between 2010 September and 2011 May, HBC 722 began to rise in brightness once more, re-achieving peak brightness by 2012 June and exceeding it in optical bands by 2013. High cadence photometric observations reveal a multitude of periods and "flickering" that may be associated with a still-active inner accretion disk (Green et al. 2013). Resolved Submillimeter Array (SMA) observations detect little remnant envelope material close to the star (Dunham et al. 2012).

SPIRE imaging was gathered in single-cycle integration times in the small map mode, with a single (349 s) observation per source. SPIRE observed simultaneously at 250, 350, and 500 μm. The on-orbit beam sizes are 181, 252, and 366, respectively. The Herschel data were reduced using HIPE (Herschel Interactive Processing Environment; Ott 2010) pipeline.

PACS imaging was acquired using two pairs of single-cycle scan maps. For each source, a "blue" (70 and 160 μm) and "green" (100 and 160 μm) simultaneous observation was taken, for two different orientation angles (70° and 110°), a total of four (279 s) observations per source. The final 160 μm image was produced by co-adding all four scans; the 70 and 100 μm images were averaged from their two respective scans each. The on-orbit beam sizes for 70, 100, and 160 μm are 55, 67, and 107, respectively.

2.3.1. SPIRE-FTS

2.3.2. PACS

2.3.3. HIFI

2.3.4. CSO

2.3.5. IRS

In order to produce a single SED for each source, we needed to calibrate the continuum flux density across Spitzer-IRS, Herschel-PACS, and Herschel-SPIRE, which range in spatial resolution from 2'' to 40''. At the shortest wavelengths of IRS, the spatial resolution was 1100 AU at a distance of 550 pc; the material emitting in the Spitzer bands remains unresolved. To determine the continuum level, we took the final output of each instrument and scaled, first between modules, and then between instruments. At shorter wavelengths with improved spatial resolution, these sources are dominated by disk/envelope warm dust within the IRS beam as listed above; the FWHM of the mid-IR continuum measured from IRS was typical of YSOs, with no more than a 20% flux mismatch between modules for the four sources observed in Green et al. (2006). Thus, in each case, we assumed that the source was essentially unresolved, and increased spatial resolution at shorter wavelengths provided a better absolute flux calibration. Where possible we then correlated this with photometry, extrapolating between missing wavelength regions. All of the scaling is described below.

The IRS data for V1515 Cyg, V1057 Cyg, and FU Ori were first presented in Green et al. (2006), and the IRS spectra for V1735 Cyg was presented in Quanz et al. (2007c); for this work, we reduced V1331 Cyg using the same technique (the SH reduction appeared in Carr & Najita 2011). "SL" (short-low module; 5.3–14 μm) data were multiplied by 1.03 and "LH" (long-high module; 20–38 μm) multiplied by 1.05 to match "SH" (short-high module; 10–20 μm) to produce the final spectrum. Additionally, we note that in V1735 Cyg, the SL spectrum was scaled by 1.35 to match SH (SL was not included in the Quanz et al. 2007c data). HBC 722 was not observed by IRS (pre- or post-outburst). As described in Green et al. (2006), the spectra were dereddened using an analytical extinction law (Mathis 1990), using an R_V of 3.1, in the cases of FU Ori, V1515 Cyg, and V1057 Cyg.

For all six sources, the PACS data were scaled, module-by-module, using a polynomial factor (as a function of wavelength), a technique that maximized continuum S/N while determining the proper absolute flux calibration. The full technique, as applied to the DIGIT data set, is described in detail in Green et al. (2013). We modified this procedure in one way: in all but HBC 722 we included all flux in the 3 × 3 spaxels as part of the FUor environment, for both continuum and lines, rather than using the full array for continuum and the smaller 3 × 3 for lines.

For HBC 722, the SPIRE flux density was much greater than the PACS continuum. This is likely because, as noted in Green et al. (2011), the continuum flux density near HBC 722 was contaminated by an embedded source, 2MASS 20581767+4353310, 17'' to the southeast. SPIRE could not resolve these sources, but Spitzer and PACS photometry could do so. We produced a PACS spectrum of the entire 3 × 3 central region summed together, which included the embedded source and was similar in spatial extent and measured flux density to that of the SPIRE beam. Thus, the composite "HBC 722" spectrum is likely dominated by the embedded source rather than HBC 722 itself. In addition, a second epoch of observations of the same field but with a different orientation angle were taken in this program, and the absolute flux matched to within 10% for both extractions. For the HBC 722 composite spectrum, we multiplied B2B by 1.5, and both R1 modules by 0.67. In all other cases, the standard DIGIT procedure (scaling to the 3 × 3 spaxels) produced a smooth PACS continuum. No scaling relative to IRS was performed, but the final PACS spectrum was a reasonable match to the extrapolated IRS continuum at 60 μm.

V1735 Cyg also suffered from confusion with V1735 Cyg SM1 (24'' to the northeast), at λ > 100 μm. The spectroscopic flux density using our standard central 3 × 3 spaxels aperture exceeded the PACS photometry by 50% at 70 and 100 μm, and by 30% at 160 μm. We infer that an annulus of ∼17'' or less is required to avoid contamination. However, the PACS spectroscopic data are not sufficiently sampled to extract the spectrum from a 17'' aperture at all wavelengths, and thus we use the larger aperture and assume contamination at these levels.

The SPIRE data were extracted using the "extended source" calibration pipeline, as this produced a smoother continuum between modules, better S/N, and fewer spectral artifacts than the "point source" pipeline. First, we multiplied SLW to match the SSW flux density in the overlap region of the two modules (300–320 μm), and then matched the flux at 210–220 μm to an extrapolation of the final PACS spectrum. HBC 722 and V1735 Cyg required no additional scaling, either between modules or between instruments. For both FU Ori and V1515 Cyg, we multiplied SSW by 0.9 to match PACS. For both V1331 Cyg and V1057 Cyg, we multiplied SLW by 1.55 to match SSW, with no additional scaling to match PACS.

For the DIGIT sample, we noted that the uncertainties in the PACS wavelength centroid were at least 30–50 km s⁻¹. With shorter exposure times in FOOSH, we found slightly more erratic centering in our line positions, up to 100 km s⁻¹. Small mispointings can masquerade as velocity shifts across the PACS IFU (see the PACS Observer Manual for further information); thus, we cannot determine the gas velocities to better than 100 km s⁻¹. The SPIRE spectra have even greater uncertainty due to lower spectral sampling and bigger beam size. Given the uncertainties, we did not observe any convincing velocity shifts in our data.

Figure 1 shows three-color (70, 100, and 160 μm) PACS images, and Figure 2 shows three-color (250, 350, and 500 μm) SPIRE images for the five classical FUors in our sample. We consider each source below. Table 3 lists the detected flux in 10''–15'' (PACS PSF corrected) and 30'' (SPIRE) apertures centered on the source coordinates. In summary, most of the FUors appear pointlike at 70 μm, but the extended filaments gradually begin to dominate at longer wavelengths.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The HBC 722 images were first presented in Green et al. (2011). All are surrounded by substantial emission from multiple sources in the far-IR which are detected within the SPIRE beam, especially at longer wavelengths. The dominant continuum source in the HBC 722 field is actually 2MASS 20581767+4353310, an embedded protostar (Green et al. 2011). They note a separate 0.41 Jy source detected at PACS 70 μm at the position of HBC 722, below the sensitivity of the spectroscopic observations.

V1515 Cyg is detected in all bands, but does not show strong continuum sources at the stellar position at SPIRE wavelengths, relative to the surrounding nebulosity. At 70 μm, the FUor is easily distinguished from the background, but at 500 μm, the arc-like emission extending 2' to the north becomes dominant.

Similarly, V1331 Cyg, V1057 Cyg, and FU Ori show clear detections with PACS but additional arc-like structures generally correlating with previously analyzed millimeter data (e.g., Kóspál et al. 2011). In each case, the FUor appears much bluer than the filamentary material.

In the case of V1735 Cyg, the optical outbursting source is the fainter western extension of the submillimeter peak; the peak emission comes from the embedded source V1735 Cyg SM1, consistent with observations by Spitzer-MIPS (Harvey et al. 2008). The beam size of MIPS at 70 μm is comparable to that of SPIRE at 250 μm, allowing marginal resolution of the two sources. This is apparent from the PACS images where the sources are spatially resolved. At longer wavelengths, they are inextricably blended.

Figure 3 shows continuum SEDs for all five FUors, including UBVR, JHK (Two Micron All Sky Survey (2MASS) and UKIDSS), Spitzer-IRS spectra (from Green et al. 2006, Quanz et al. 2007c, and this work), and PACS/SPIRE photometry (this work). For comparison, we also show the incomplete SED of HBC 722 from Green et al. (2011); the PACS/SPIRE spectra for this source are clearly contaminated, as evidenced by the mismatch with the 70 μm PACS photometric data point, and the non-detection at 1.3 mm with the SMA (Dunham et al. 2012).

Zoom In Zoom Out Reset image size

Additionally, the PACS photometry falls below the PACS spectrum for V1735 Cyg and V1515 Cyg. The former can be explained by contamination in the spectra from V1735 Cyg SM1, which is cleanly separated by the photometry but not by the spectroscopy. A similar situation is likely in V1515 Cyg, where the spectroscopic data include additional surrounding nebulosity.

The most notable slope discontinuity between PACS and SPIRE is in the FU Ori SED; the emission at SPIRE wavelengths is faint compared to the other sources. In V1331 Cyg, the longest wavelength SPIRE photometry, at 500 μm, seems contaminated by extended emission. In the other, brighter sources, the PACS and SPIRE photometry and spectroscopy are very well aligned.

As FUors are variable and generally fade over time, we used photometry as close to contemporaneous as possible. We then calculated L_bol and T_bol using the technique from Dunham et al. (2010). In summary, T_bol is the temperature of a blackbody with the same flux-weighted mean frequency as the measured spectrum and L_bol is the total integrated luminosity, using trapezoidal interpolation between data points.

3.2.1. Classification

The original definition of Class I and II sources for YSOs is derived from the ratio of the 2 μm/25 μm flux density (Andre et al. 1993; Greene et al. 1994), the spectral index α. For HBC 722, there is a pre-outburst measurement of α at −0.77, which would fall under the definition of Class II (Miller et al. 2011). Quanz et al. (2007c) used mid-IR spectral indices and found the FUors to be indistinguishable from Class II objects in Taurus (Furlan et al. 2006). We compare the classifications of the non-confused FUors in our sample using three different classification methods, explained below, in Table 4.

Table 4. Classification

Object	α	Tbol	Env. Mass
	Class	Class	Stage
V1057 Cyg	FS	II	I
V1515 Cyg	II	II	I
V1331 Cyg	II	FS	I/IIa
FU Ori	II	II	II
V1735 Cyg	II	FS	–
HBC 722	IIb	–	–

Notes. Classification of the FOOSH sources by three different techniques: mid-IR spectral index α, T_bol, and envelope mass (see the text). "FS" indicates a flat-spectrum classification. Contaminated measurements are reported as "–." ^aV1331 Cyg can be classified as Stage I or II depending upon the wavelength used to measure the envelope mass; the inferred mass is close to the boundary of 0.10 M_☉ defined in Crapsi et al. (2008). ^bα was measured pre-outburst for HBC 722 (Miller et al. 2011).

Download table as:  ASCIITypeset image

In general, T_bol can be used as a proxy for the evolutionary state (Chen et al. 1995; Robitaille et al. 2006): in this framework, a Class I source has T_bol between 70 and 650 K and a Class II source has T_bol > 650 K. Flat-spectrum sources can be characterized as having 350 K < T_bol < 950 K (Evans et al. 2009; Fischer et al. 2013). T_bol is quite sensitive to the near-IR flux and is thus highly uncertain in variable sources, such as FUors. With that caveat in mind, we find that FU Ori, V1515 Cyg, and V1057 Cyg exhibit Class II SEDs, while V1735 Cyg and V1331 Cyg exhibit typical flat-spectrum SEDs near the Class I/II boundary. This sequence agrees with the sequence from SED color, extinction, and strength of silicate emission found in Green et al. (2006) and Quanz et al. (2007c). The lone surprise is V1331 Cyg which might be expected to be the second-most blue/evolved source, given its silicate emission feature and flat/slightly declining mid-IR SED. In summary, the T_bol values suggest that FUors are more evolved than Class I sources. Unfortunately, the value of T_bol as an evolutionary proxy is debatable. In the case of an FUor flare, models suggest that the far-IR luminosity will increase within a few months of outburst (Johnstone et al. 2013), although Ábrahám et al. (2004) find, in contrast, that changes to the SED occur primarily in the optical/IR rather than the submillimeter on the 1–10 yr timescale. If the far-IR lagged significantly behind, then this would increase T_bol without any change in the submillimeter SED or derived envelope mass. However, this scenario would require the dust to be 1–10 lt-yr (0.3–3.0 pc) distant, and hence unrelated to the FUor.

Thus, we also consider a criteria related to the physical stage of the object, using the Herschel spectra. We calculate the envelope mass using the "OH5" opacity law (Ossenkopf & Henning 1994) as described in, e.g., Dunham et al. (2011). First, we determine an approximate dust temperature from the shape of the PACS and SPIRE continuum from 100 to 600 μm. We compare dust at this temperature, assuming an emissivity α λ^−1.8, normalized to the 350 μm continuum, to confirm that the model slope is similar to the observed spectrum. Then, we calculate an envelope mass from the SPIRE photometry. The results for the four isolated sources in our sample are shown in Table 5. Using this criteria, Crapsi et al. (2008) consider 0.1 M_☉ as the envelope mass cutoff between Stage I and II sources. Their boundary value was derived from Robitaille et al. (2006) models, converting their boundary accretion rate value of 10⁻⁶ M_☉ yr⁻¹ to 0.07 M_☉ yr⁻¹, with the additional consideration that disks greater than 0.1 M_☉ were gravitationally unstable. However, there are two dominant sources of uncertainty in this calculation. First, the dust temperature is not perfectly constrained and varies between 38 and 54 K, causing an uncertainty in the derived envelope mass of ∼35%. Second, the estimated mass increases significantly if derived using longer wavelengths, due to the inclusion of increasingly larger areas of colder dust from the more distant parts of the envelope, suggesting that values calculated at 350 μm generally underestimate the envelope mass.

Table 5. Dust-derived Envelope Mass

Object	$\mbox{$L_{\rm bol}$}$/$\mbox{$L_{\rm smm}$}$	Tdust	Env. Mass (250 μm)	(350 μm)	(500 μm)
		(K)	(M☉)	(M☉)	(M☉)
V1057 Cyg	975	43	0.16	0.25	0.30
V1515 Cyg	505	42	0.13	0.19	0.28
V1331 Cyg	159	38	0.11	0.07	0.46
FU Ori	5858	54	0.02	0.05	0.07

Notes. Ratio of bolometric over submillimeter luminosity, dust temperature, and envelope mass calculated at each of the three SPIRE photometric bands. The large envelope mass derived at 500 μm in V1331 Cyg is attributed to extended/diffuse emission.

Download table as:  ASCIITypeset image

The results for each classification method are summarized in Table 4. The α and T_bol classifications agree very well, while the envelope mass method indicates an earlier evolutionary stage than the other methods in the cases of V1057 Cyg and V1515 Cyg. If we crudely average the classifications of the three different methods (assigning values of 1, 1.5, and 2 to Class/Stage I, flat spectrum, and Class/Stage II, respectively), then we find the following sequence in increasing order of Class/Stage: V1057 Cyg (1.5), V1515 Cyg (1.67), V1331 Cyg (1.67), V1735 Cyg (1.75), and HBC 722 (2.0). This ordering is broadly in agreement with the mid-IR dust feature sequence from Quanz et al. (2007c). Therefore, we find that the umbrella group of "FUors" includes both Stage I and II objects.

3.2.2. Bolometric Luminosity

We do not observe any evidence for cold crystalline Mg-rich or Fe-rich dust at 69 μm, which has been seen in some T Tauri and Herbig Ae/Be star disks (Sturm et al. 2010; Sturm et al. 2013). However, our sensitivity is substantially lower than in sources with detections of this feature, due to shorter integration times and greater distances to the source (lower S/N). Furthermore, the sources in Sturm et al. (2013) typically show crystalline silicate dust in the IRS bands, whereas all of the FUor IRS spectra are pristine (Quanz et al. 2007c). At this time, calibration uncertainties in spectral shape prevent a search for 60–70 μm H₂O crystalline ice features.

For the DIGIT embedded objects, we used separate extractions to determine line spatial extent as distinct from continuum (Green et al. 2013). This was unnecessary for the FOOSH sources, which show no evidence of extended line emission, except in cases of contamination. Differences in the continuum and line extent resolved with Herschel in FUors is attributed to large-scale structure, multiplicity, or outflows and is discussed in the following section. For this work, we follow a slightly simpler procedure than that used in Green et al. (2013), and assume the line and continuum emitting regions are the same, and apply the same scaling factors.

The list of detected lines in PACS and SPIRE for all six source regions is reported in Table 6. In summary, the PACS observations show few lines overall, while the SPIRE spectra are richer. The only lines detected in all six sources are [O i] 63 μm and [C i] 370 and 610 μm. Other observed fine structure lines include [O i] 145 μm, [N ii] 205 and 122 μm, and [C ii] 158 μm. We detect CO J = 5 → 4 and J = 4 → 3 in all but FU Ori (although CO J = 5 → 4 is detected only in the HIFI data, below the detection threshold of SPIRE due to the narrow feature width). CO J = 7 → 6 and higher are seen only in V1057 Cyg, V1735 Cyg, and HBC 722; in V1057 Cyg, we detect CO up to J = 23 → 22. We detect ¹³CO J = 8 → 7 to J = 5 → 4 in V1735 Cyg and HBC 722. We detect H₂O 174.6 μm, and tentatively detect OH 84.41 μm, in one source (V1057 Cyg). OH 119 μm, the lowest state of the OH 3/2 → 3/2 ladder, is the only line in absorption (seen in V1735 Cyg).

None of the sources were detected in CO J = 14 → 13 with HIFI, but all were detected in CO J = 5 → 4. Figure 4 (top) shows the CO J = 5 → 4 HIFI (black) compared with the HCO⁺ J = 3 → 2 (267.558 GHz; blue) line observed with the CSO. FU Ori was not observed in HCO⁺ in our observing run; all of the others—V1515 Cyg, V1735 Cyg, V1057 Cyg, V1331 Cyg, and HBC 722—were observed to emit in the HCO⁺ J = 3 → 2. In the case of HBC 722, we also observed CO J = 4 → 3 and J = 2 → 1. The line profiles are compared in Figure 4 (bottom), placed on the same vertical scale for comparison. (Note that the CO integrated intensity is scaled by the factors listed in each subfigure.)

Zoom In Zoom Out Reset image size

As an example, in Figure 5, we present the 50–670 μm continuum-subtracted spectrum of V1057 Cyg (top), rebinned to low resolution for clarity, versus V1735 Cyg (bottom), which is partly contaminated by V1735 Cyg SM1. In this form, a few details become apparent. The peak CO line flux appears at much longer wavelengths in V1735 Cyg, compared to V1057 Cyg or to a typical embedded source (e.g., Figure 15, Green et al. 2013). The faint ¹³CO ladder appears from J = 5 → 4 to J = 8 → 7 only in the case of V1735 Cyg. The only detection of H₂O (at 174.6 μm) is in V1057 Cyg (the stronger nearby line is CO J = 15 → 14 at 173.6 μm). V1735 Cyg shows a relatively weak [O i] feature in comparison to the CO.

Zoom In Zoom Out Reset image size

The remaining FUor spectra (including the contaminated HBC 722 from Green et al. 2011) are shown in Figures 6 and 7.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Although we detect substantial line emission from three sources (V1057 Cyg, V1735 Cyg, and HBC 722), we must consider whether the lines are local to each FUor, using the spatial information provided by the PACS array. It is more difficult to determine this with SPIRE due to the larger beam, sparse spatial coverage, and lack of off-position data.

3.5.1. PACS

Figures 8 and 9 show 10% contours for all detected lines (gray scale) and local continuum (red) with PACS. The local continuum is selected from a line-free set of channels surrounding each line. In HBC 722, we see a clear demarcation between the FUor (central spaxel) and the embedded source (two spaxels to the SE). The submillimeter continuum is strongly dominated at all wavelengths by the embedded source. For V1735 Cyg, V1735 Cyg SM1 appears as a continuum source cleanly detected two spaxels to the east, at the edge of the array. In the other four sources (V1515 Cyg, V1057 Cyg, V1331 Cyg, and FU Ori), the central spaxel dominates the continuum emission, and the pointing accuracy can be estimated as within 0.2 spaxels from the centroid at each wavelength. The continuum footprint grows with wavelength, consistent with the point-spread function (Green et al. 2013); thus, there is no evidence for extended PACS continuum emission in these four sources.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The line emission differs from the continuum profiles in a few cases. In the HBC 722 composite, the [O i] 63.18 μm emission is cospatial with the FUor, while the CO J = 16 → 15 emission extends along both sources in an SE–NW band roughly correlated with the CO J = 2 → 1 map (Green et al. 2011). The marginally detected H₂O emission at 179.53 μm is associated with the embedded object to the SE. We attribute the [O i] to the FUor, although we note that the [O i] could be excited by shocks driven by other millimeter sources identified in SMA observations (Figure 1, MMS1 and MMS2; Dunham et al. 2012). In V1735 Cyg, the [O i] emission is equally prominent around both the FUor and the SM1 source, but CO J = 16 → 15 is compact and associated only with SM1. In V1057 Cyg, V1515 Cyg, V1331 Cyg, and FU Ori, the [O i] and CO emission is compact and aligned with the FUor when detected (although there is a hint of extended emission in FU Ori [O i]). There is very little line emission detected in V1515 Cyg at all. [N ii] 122 μm is detected in off-center spaxels in V1515 Cyg, suggesting that the emission is diffuse. [O iii] (88.35 μm) and [N iii] (57.0 μm) appear only in absorption, indicating emission lines in the off-position. In HBC 722, the rotation of the off-position between our two data epochs dominated change in the flux of the [C ii], [O iii], and [N iii] absorption features, which is clear evidence that they are derived from the off-position and are not associated with the sources in the PACS field.

3.5.2. SPIRE

3.5.3. Summary of Results

HBC 722 and V1735 Cyg are the sources with the strongest [O i] 63 μm lines. They are also the only two sources with [O i] 145 μm detections. The 63/145 μm [O i] flux ratios are 21 and 11, respectively. The ratios fall slightly outside the range of the DIGIT embedded sample (14–20) for HBC 722 and V1735 Cyg. For the other sources, the lower limit to the 63/145 μm ratio is 15 (V1057 Cyg), 7.6 (FU Ori), and 1.5 (V1331 Cyg and V1515 Cyg).

Figure 10 shows the distribution of [O i] 63 μm line luminosity for the DIGIT embedded and FOOSH samples. A K-S test shows that the FUors are significantly different from the protostars as a group. The average (median) [O i] luminosity is 4.3 (3.1) × 10⁻³ L_☉ for the FUors, compared to 1.6 (0.5) × 10⁻³ L_☉ for the Class 0/I sources. If we compare to more evolved sources, then the strongest [O i] emission in any Herbig Ae/Be star in the DIGIT sample is 1.4 × 10⁻³ L_☉, and most Herbig/Class II sources fall below 10⁻⁴ L_☉ (Meeus et al. 2013; Fedele et al. 2013) for cases where the [O i] line is detected at all.

Zoom In Zoom Out Reset image size

Figure 11 shows the correlation between [O i] line strength and L_bol across the DIGIT embedded and FOOSH samples. The large [O i] line strength in the FOOSH sample can be attributed to greater L_bol, albeit with considerable scatter.

Zoom In Zoom Out Reset image size

We consider whether shocks contribute significantly to [O i] emission in our sample. The [O i] can be excited by either UV radiation from the central star or shocks (Hollenbach 1985). The earliest outburst in the FOOSH sample, FU Ori, occurred 75 yr prior to the Herschel observations. At 300 km s⁻¹, the wind-lofted material would have propagated ∼4800 AU, slightly less than the beam size of ∼6000 AU for these sources. The emitting area for [O i] increases over time as the shock propagates outward, so a propagating shock model might predict higher [O i] line fluxes from the older outbursts, which we do not observe.

Moreover, a slower wind speed of 100 km s⁻¹ is more consistent with the accretion rate in five of the six sources. The accretion luminosity is calculated (e.g., Hartmann 1998):

$$\begin{equation} L_{{\rm acc}} = \frac{GM\mbox{$\dot{M}$}}{2\ R_{\star }} \end{equation} \tag{ 1 }$$

where R_⋆ is taken as 2 R_☉. The [O i] line luminosity is related to the instantaneous accretion rate, $\mbox{$\dot{M}$}=$10⁴ L([O i] 63 μm, in L_☉), in units of 10⁻⁵ M_☉ yr⁻¹ (Hollenbach 1985). We compare these two mass-loss indicators in Figure 12. The mass-loss rates are consistent with a wind speed of 100 km s⁻¹, except in the case of HBC 722, in which the accretion luminosity underestimates the [O i] luminosity. The shock had only ∼ 3 months to propagate since peak outburst (∼5 AU at a shock speed of 100 km s⁻¹), and thus the physical region of [O i] enhancement due to shocks should be small, yet we observe the opposite effect (albeit at weak significance).

Zoom In Zoom Out Reset image size

In contrast, the [O i] line flux from UV radiation should track L_bol. Instead, the mass-loss rate indicated by [O i] is constant as a function of accretion luminosity, between ∼10⁻⁵ and 10⁻⁶ M_☉ yr⁻¹ for all six sources. This suggests that either the [O i] emission is dominated by the current UV field, or that contributions from outflow-driven [O i] emission predate the current outburst. This is particularly interesting given that HBC 722 appears to be a relatively evolved source with little remnant envelope to fuel multiple burst cycles at this stage.

In addition to [O i], CO is detected in a few transitions in FU Ori, V1515 Cyg, and V1331 Cyg, and in 20 transitions in V1057 Cyg. One useful tool in analyzing molecular emission is the rotational diagram; a detailed review can be found in Goldsmith & Langer (1999), and a brief review in the context of Herschel spectroscopy is in Green et al. (2013). As with samples of protostars, we can fit components to distinct ranges of CO J_up: to reduce artifacts due to absolute flux calibration, we consider separately CO from SPIRE (J_up ⩽ 13), PACS R1 (14 ⩽J_up ⩽ 24), and PACS B2A/B2B (25 ⩽J_up; no detections in this sample). The results are presented in Figure 13.

Zoom In Zoom Out Reset image size

Most Class 0/I (embedded and non-outbursting) sources observed with Herschel-PACS exhibit CO emission up to J = 24 → 23 or even higher (Manoj et al. 2013; Karska et al. 2013; Green et al. 2013). These sources consistently show a 300–400 K component in this frequency range (van Kempen et al. 2010), attributed to UV irradiation of the outflow cavity walls and C-shocks (Visser et al. 2012). This component is usually referred to as "warm." Many embedded sources also exhibit a "hot" component detected in CO transitions J_up ⩾ 25, ranging from 700 to 900 K. We use the same definitions in this work, and add the "cool" component to identify fits to CO transitions J_up ⩽ 13; the source of the cool emission is usually attributed to passive heating (from gas particle collisions with dust grains heated by stellar radiation; Visser et al. 2012).

We do not observe a warm component in most of these systems. Of the FUors, only V1057 Cyg shows robust cool and warm components at 81 and 368 K, respectively. Lorenzetti et al. (2000) noted a detection of CO J = 17 → 16 for V1057 Cyg only, but no higher- or lower-J transitions, using ISO-LWS. We also fit cool components to V1331 Cyg and V1515 Cyg and find T_rot of 30 and 19 K, respectively. We detect only one CO line in this range in FU Ori (J = 5 → 4) from HIFI, and thus cannot compute T_rot.

We also fit a single component to the CO and ¹³CO rotational diagrams for V1735 Cyg (Figure 14). We find T_rot of 67 K and 45 K, respectively, but do not attribute the emission solely to the FUor.

Zoom In Zoom Out Reset image size

As the highest observed rotational level of CO emission decreases, the rotational temperature of gas decreases as well, from 81 K in V1057 Cyg to 19 K in V1515 Cyg. The total amount of CO gas, $\mathcal {N}$(CO), dominated in the FUors by the cool CO, does not follow this trend but instead varies more like L_bol. This is consistent with $\mathcal {N}$(CO) derived from the warm component in embedded sources (Manoj et al. 2013; Karska et al. 2013; Green et al. 2013). If the gas around V1057 Cyg is well approximated by a two-temperature (cool/warm) fit, then optical depth effects would be most apparent at low-J, where the line flux would fall below the fit for certain density regimes (Goldsmith & Langer 1999, Figure 12). Instead, the CO continues to exhibit positive curvature to the longest SPIRE wavelengths. This suggests that either the gas is populated at low densities (Neufeld 2012; Manoj et al. 2013), that optical depth effects are not substantially affecting the V1057 Cyg line fluxes even at low-J, or that there is a larger reservoir of cooler gas.

The best constraint on opacity arises from the ratio of ¹³CO to ¹²CO: in V1057 Cyg, the upper limit of the isotopic ratio of the J = 5 → 4 line intensity is ∼10, requiring an optical depth less than ∼7. For comparison, the CO optical depth measured from the ¹²CO/¹³CO ratio declines with upper state J for both confused sources V1735 Cyg and HBC 722, from τ of ∼7 at J = 5 → 4 down to ∼3 at J = 7 → 6.

In the limit of moderate or higher density, the observed curvature suggests that additional "cold" components, or even a power-law distribution of components, may be needed to fit the rotational diagram in order to produce sufficient low-J emission. It has already been suggested that a power-law distribution is a good fit to the CO J = 14 → 13 up to J = 49 → 48 PACS lines detected in the Orion protostar sample (Manoj et al. 2013); whether or not this applies to the FUors is not clear.

4.2.1. Velocity-resolved Mid-J Emission

4.2.2. Where is the Hot CO Emission in FUors?

We might posit that CO emission from hot gas would be a transitory signature of heating from a recent outburst. How then do we explain the lack of high-J CO emission in these FUors?

Analysis of the non-detections reveals that the absence of high-J CO and all but one H₂O line in our sources is of only modest significance. Figure 15 shows histograms of the line luminosities of CO J = 16 → 15 (left) and H₂O 174.63 μm (right) for the DIGIT sample (Green et al. 2013). The vertical dashed lines show the strength of these lines in V1057 Cyg; the vertical dotted lines show 3σ upper limits for the non-detections in the FOOSH sample. In both cases, the detected CO J = 16 → 15 line (in V1057 Cyg) would be among the brightest in the DIGIT protostar sample, but not a notable outlier. The upper limits from our sample for CO and H₂O are not low enough to rule out significant emission in any source.

Zoom In Zoom Out Reset image size

If the CO emission originates in a tenuous envelope, then there may not be enough emitting material to produce a detectable signature in the warmer regions. If the CO emission originates in a disk, then we require a thermal inversion to produce emission lines in the upper layer. Models suggest that the midplane of FUors may become hotter than the upper disk layers during outburst, which would mask the signature of CO (Zhu et al. 2010). Our data do not sufficiently constrain the line luminosities to distinguish the FUors from ordinary protostars.

In the DIGIT sample, the CO emission from disks is much weaker. The highest line luminosity from CO J = 16 → 15 detected in the DIGIT disks sample is in HD 100546, a Herbig Ae/Be star of similar luminosity to the FUors (Sturm et al. 2010). The CO line is a factor of 20 weaker in luminosity than V1057 Cyg, at 3 × 10⁻⁵ L_☉ (Meeus et al. 2013). Thus, it is unlikely that the disk is a significant contributor to the CO line luminosity in V1057 Cyg.

We consider here a few other marginal detections. We report the detection of a single line of H₂O at 174.63 μm in V1057 Cyg. The H₂O line luminosity in V1057 Cyg is 0.47 × 10⁻³ L_☉, which would place this single line among the most luminous of the DIGIT sample. The 174.63 μm line (excitation energy of 197 K) is not even the lowest excitation line within the PACS bands; in embedded sources, the lowest energy detected H₂O line is at 179.53 μm (114 K); however, the 174.63 μm line is typically among the most luminous H₂O lines detected in Class 0/I protostars. We also do not detect the 557 GHz 1₁₀ − 1₀₁ line in the SPIRE bands. We do not detect H₂O in any other sources. Although the detection in V1057 Cyg is robust, the upper limits on this line in the other FUors do not rule out substantial emission lines as well.

We tentatively detect OH 84.41 μm, half of a doublet in the 3/2 → 3/2 ladder, also in V1057 Cyg. Although this is the most commonly detected line in the embedded sample, the 84.61 μm half of the doublet is typically just as strong. More prominently, OH 119 μm is seen in absorption in V1735 Cyg. Only V1735 Cyg and 2 of 30 Class 0/I sources in the DIGIT sample show this line in absorption. There is also a hint of OH emission at 163 μm, possibly blended with a weak detection of the CO J = 16 → 15 line, but both detections are inconclusive.

No other molecules are detected in the FOOSH sample with Herschel. In contrast, at least some embedded sources show rich SPIRE spectra (Goicoechea et al. 2012; J. D. Green et al., in preparation). However, the upper limits for H₂O and OH line luminosity are consistent with the detections in embedded sources.

[C ii] 157.74 μm is detected in V1735 Cyg, V1331 Cyg, and V1515 Cyg, but is not clearly associated with the sources, similar to the DIGIT embedded sources.

[C i] is observed at 370 and 610 μm in all sources. The [C i] 370/610 μm flux ratio falls between 2.1 and 3.0 across the sample, consistent with ISM values (Jenkins & Tripp 2001), and the [C i] emission is associated with cooler extended gas rather than the FUors themselves. The SPIRE maps show that the lines are extended (at beam size ∼40''); we do not attribute the emission to the FUors.

[N ii] 205 μm (SPIRE) is observed to emit strongly in HBC 722, V1331 Cyg, V1057 Cyg, and V1515 Cyg, but not V1735 Cyg or FU Ori. The [N ii] 122 μm (PACS) line is detected along with the 205 μm line around V1515 Cyg; no other sources in our sample show the 122 μm line. This is in contrast with ISO-LWS results, in which [N ii] was marginally detected at 122 μm with ISO-LWS in V1735 Cyg, V1331 Cyg, and V1057 Cyg (Lorenzetti 2005). As noted earlier, the difference in the detection rates of [N ii] 122 μm is likely due to the use of an off-position for our PACS data but an on board calibration source for SPIRE.

We detected both lines around V1515 Cyg, and examined the PACS and SPIRE maps to see if there was any hint of centrally peaked emission. The flux of each line shows no significant variation (< 50%) across the entire PACS and SPIRE field of view, and is uncorrelated with proximity to the central position. The [N ii] 122/205 μm flux ratio is diagnostic of the electron density (Hudson & Bell 2004; Oberst et al. 2011). The ratio in the central pixel is 0.31, suggesting an extremely low electron density (n_e < 1 cm⁻³). We note that the observed ratio should be considered an upper limit, as the PACS spectra are observed using an off-position for sky subtraction, while the SPIRE spectra are not.

Further evidence of the diffuse nature of the [N ii] emission comes from the fact that there is no clear trend between [N ii] 205 μm and the strength or position of the CO emission. Thus, we conclude that among fine structure lines, only the [O i] emission is localized to the FUor.

It has been posited from occurrence rate (Hartmann & Kenyon 1996) and from simulation (e.g., Dunham & Vorobyov 2012) that FUors are Stage 0/I stars undergoing episodic outburst cycles with replenished material from an infalling envelope. However, FUors have much in common with Class II sources: in Spitzer-IRS bands, all but V1735 Cyg show silicate emission and optical-IR continuum indices consistent with relatively low extinction (e.g., Hartmann & Kenyon 1996; Green et al. 2006; Quanz et al. 2007c). It has been suggested that the overall group of FUors falls into two different categories, those consistent with flared disks and those requiring envelopes (e.g., Kenyon & Hartmann 1991; Zhu et al. 2008).

Zhu et al. (2008) modeled the dust from Spitzer-IRS observations of FU Ori, V1057 Cyg, and V1515 Cyg and found that V1057 Cyg and V1515 Cyg required envelopes typical of protostars, while FU Ori did not require an envelope. The Herschel spectra generally fall above their model predictions, indicating an underestimate of the reservoir of cold dust surrounding these systems. Nonetheless, our estimates of the envelope mass from the Herschel dust emission broadly agrees with this classification; we find that V1057 Cyg and V1515 Cyg have envelopes greater than 0.1 M_☉, while FU Ori has a much smaller envelope mass of ∼0.02 M_☉. HBC 722 was considered a T Tauri star prior to outburst, and measurements of the envelope mass upper limit suggest a tenuous envelope at best.

We noted some inconsistencies between the different classification methods for the continuum, but found broad agreement with the mid-IR sequence in Quanz et al. (2007c). However, FUors may not be well characterized by the Stage I/II sequence because the Herschel line observations provide a different picture from the continuum. In general, the presence of weaker silicate features and larger mid-IR α is correlated with increased submillimeter dust and CO/H₂O emission, but is not well correlated with [O i]. Class 0/I sources typically exhibit hotter and more excited CO rotational lines than Class II sources (cf. Karska et al. 2013, Meeus et al. 2013); by this standard, FUors would seem to resemble disk sources, however, these lines may be too faint to detect in our observations. The DIGIT (and HOPS) samples show a tight correlation between CO J = 16 → 15 line luminosity and derived $\mathcal {N}$(warm). In V1057 Cyg, the only FUor where the warm gas is detected, it falls in the DIGIT embedded source correlation. Additionally, the DIGIT sample showed a lack of correlation between T_rot(warm) and $\mathcal {N}$(CO); again, the lone FUor (V1057 Cyg) is consistent with the embedded sample. Finally, the DIGIT sample showed a (weaker) trend between L_bol and $\mathcal {N}$(warm); once again, V1057 Cyg fits this trend. It is clear that the spectra of V1057 Cyg, as characterized by PACS, would not distinguish it as an outlier in the DIGIT Class 0/I sample, even though it has a Class II SED, suggesting it is a kind of hybrid object. However, the high upper limits on line emission do not provide strong constraints on the evolutionary state of the other FUors.