THE LATE STAGES OF PROTOPLANETARY DISK EVOLUTION: A MILLIMETER SURVEY OF UPPER SCORPIUS^*

The spectral types of BA stars are from Hernández et al. (2005), and for KM stars we use spectral types from Preibisch et al. (1998, 2002). For BA stars, we adopt stellar masses from Hernández et al. (2005), who used the stellar isochrones of Palla & Stahler (2000). For KM stars, we use masses from Kraus & Hillenbrand (2007), who used the models of D'Antona & Mazzitelli (1997) and Baraffe et al. (1998) for solar-mass and low-mass stars, respectively. Two stars lack a mass estimate in the literature, [PBB2002] USco J160823.2-193001 and [PBB2002] USco J160900.0-190836; we assigned these stars masses according to their spectral type, as in Kraus & Hillenbrand (2007). Comparison of the KM star masses to other estimates in the literature (e.g., Preibisch & Zinnecker 1999) suggests a typical stellar mass uncertainty of 10%–20%. Both spectral type and estimated stellar mass are listed in Table 1.

The Hα equivalent width, W_λ(Hα), was taken from Hernández et al. (2005), Preibisch et al. (1998, 2002), Dahm & Carpenter (2009), and Riaz et al. (2006). This is then used to determine the accretion state of the stars. Traditionally, KM stars with W_λ(Hα) of 10 Å or greater were considered to be accreting, and referred to as Classical T-Tauri stars (CTTS). Those with lower or undetected emission were considered to be non-accreting, and referred to as Weak-line T-Tauri stars (WTTS). Recent work has refined this classification system (Barrado y Navascués & Martín 2003; White & Basri 2003), identifying spectral type dependent W_λ(Hα) that unambiguously indicate accretion. Here, we adopt the system of White & Basri—W_λ(Hα) in emission of 3, 10, 20, and 40 Å indicate accretion for K0–K5, K7–M2.5, M3–M5.5, and M6 and later spectral types, respectively. While there are a few accreting systems among the KM stars of Upper Sco, none of the BA stars in Upper Sco show Hα in emission, suggesting that none of these stars still experience accretion.

We note that the use of an accretion indicator to demonstrate the presence of gas in the disk is subject to two main problems: accretion and photospheric Hα emission can be difficult to distinguish at low equivalent widths, and accretion may vary by large amounts on timescales as short as days (Bouvier et al. 2007; Nguyen et al. 2009). Among the 20 KM stars in our sample with multiple epochs of W_λ(Hα) measurements, two show values which would produce different classifications as CTTS or WTTS over a decade timescale. In these two cases ([PBB2002] J160545.4-202308 and [PBB2002] J160702.1-201938), we classify the sources as CTTS. We list the W_λ(Hα) and accretion state of our sources in Table 1.

Our observations include 8 of the 10 CTTS in the original sample of Carpenter et al. (2006), and 17 of the 95 WTTS. There are an additional 22 KM stars with no Hα measurement in the literature, of which we include 1. The millimeter-detected sources we include from the literature are split between accreting and non-accreting systems; 1 is a CTTS and 2 are WTTS.

Table 1 also shows the disk classification based on SEDs. SEDs were constructed using Two Micron All Sky Survey J-, H-, and K-band photometry from Skrutskie et al. (2006), 4.5, 8.0, and 16 μm photometry from Carpenter et al. (2006), and 24 and 70 μm photometry from Carpenter et al. (2009). We used extinction estimates from the literature (Hernández et al. 2005; Preibisch et al. 1998, 2002) and the R = 3.1 reddening law of Fitzpatrick (1999) to deredden NIR photometry.

For the BA stars in Upper Sco, we use the SED classifications of Carpenter et al. (2009). Among BA stars, they found only debris disks. Carpenter et al. classified stars in Upper Sco as hosting debris or primordial disks on the basis of the strength of the NIR excess and comparison of NIR colors with those of disks in the younger IC348 region.

For KM stars, the classification is based upon the slope of the SED at NIR wavelengths, as in Greene et al. (1994). Assuming λF_λ∝λⁿ, sources exhibiting a near-infrared power-law index greater than 0.3 are considered Class I, −0.3 < n < 0.3 are flat-spectrum, −1.6 < n < −0.3 are Class II, and n < −1.6 are Class III. The classes roughly trace the expected evolution of inner disk dust. We use the broadest baseline applicable to our entire sample, calculating the index from K-band and 24 μm photometry.

We classify the K2 star [PZ99] J160421.7-213028 as hosting a transition disk, based on its lack of NIR excess at wavelengths less than 16 μm and steeply rising excess at longer wavelengths (e.g., Cieza et al. 2007).

The Upper Sco sample of Carpenter et al. (2006) contains 19 Class II sources, comprising 15% of the KM stars, and 107 Class III sources. We observed 16 Class II and 9 Class III sources. The three additional detections included from the literature are Class II sources.

Due to the variety of disk types among KM stars, we summarize their disk classifications in Table 2, where we show the number of each type of object in the parent sample of Carpenter et al. (2006) (applying our Class II/III and CTTS/WTTS criteria as described above), the number included in this work, and the number of millimeter detections.

Dust masses were estimated from the 1.2 mm flux density assuming the dust to be optically thin, and are listed in Table 3. Under this assumption, the dust mass (M_dust) will be directly proportional to the flux (F_ν), as in Hildebrand (1983):

$$\begin{equation} M_{{\rm dust}} = \frac{d^2F_{\nu }}{\kappa _{\nu }B_{\nu }(T_c)}. \end{equation} \tag{ 1 }$$

Assuming a dust mass opacity (κ_ν) of 10 cm²/g at 1000 GHz and opacity power-law index β = 1 (Beckwith et al. 1990), as well as a characteristic dust temperature (T_c) of 20 K (the median temperature of Taurus disks; Andrews & Williams 2005), we can then estimate dust masses as M_dust = 1.5 × 10⁻³F_1.2mm(mJy) M_Jup at the 145 pc distance of Upper Sco. For the two objects observed at 1.3 mm, [PBB2002] USco J160823.2-193001 and [PBB2002] USco J160900.7-190852, we must adopt a different constant: M_dust = 1.9 × 10⁻³F_1.3mm(mJy) M_Jup. While there is considerable theoretical uncertainty in these terms, the values are consistent with previous disk studies. The addition of far-infrared photometry will allow for detailed modeling of the dust structure of the disk and an improved dust mass estimate (e.g., Pinte et al. 2006), while multi-line spectroscopy is necessary to constrain the gas mass (e.g., Kamp et al. 2011). These are the subject of a future paper.

Our median 3σ sensitivity of 2.8 mJy corresponds to an estimated dust mass sensitivity of 4.2× 10⁻³ M_Jup. In order to estimate an upper limit to the average dust mass for non-detected BA and KM stars, we stack our non-detections. If many sources have mm-emission just below our 3σ limits, then we would expect stacking to result in a measurable flux. However, this procedure results in a non-detection for both the stacked BA and KM observations. The 3σ upper limits of the stacked observations, 0.89 mJy for BA stars, and 0.73 mJy for KM stars, do not differ from what would be expected from the noise in a single observation of the combined observation times. These upper limits suggest that on average, BA and KM stars have dust masses less than 1.3× 10⁻³ and 1.1× 10⁻³ M_Jup, respectively.

The millimeter-detected KM stars exhibit Hα emission similar to the non-detected KM stars. For each star, we calculate the mean Hα equivalent width from values in the literature, and show these in comparison to the dust masses in Figure 2. We show sources with Hα in absorption as having an upper limit to emission of 0.1 Å. We calculate the Kendell's rank statistic, generalized for a censored data set (Isobe et al. 1986), for the KM stars and find a Z-value of 0.028, indicating a 98% chance there is no correlation between the Hα equivalent widths of Upper Sco KM stars and their disk dust masses. Considering just the mm-detected sources, there is ambiguous evidence for an anti-correlation (Z-value of 1.05, 29% chance of no correlation).

Zoom In Zoom Out Reset image size

To carry out detailed comparison of the SED shapes, we plot the SED indices from 2.2 to 8 μm and 8 to 24 μm in Figure 3, defining the SED index $n_{1\hbox{--}2} = \log ({\lambda _2F_{\lambda _2}}/{\lambda _1F_{\lambda _1}})/\log ({\lambda _2}/{\lambda _1})$ as in Furlan et al. (2006). The central cluster of points correspond to Class II sources, while points along the left-hand side show no excess at 8 μm, indicative of a void inner region. However, some of these sources exhibit 24 μm excesses indicative of some amount of cooler dust which may either represent transition or debris disks. The majority of millimeter-detected sources in Upper Sco exhibit NIR SEDs typical of other Upper Sco Class II sources.

Zoom In Zoom Out Reset image size

While the NIR and Hα properties of the majority of millimeter-detected sources do not stand out among Upper Sco source, disk dust mass does appear as if it could be correlated with stellar mass. In Figure 4, we show the dust mass as a function of stellar mass. The highest dust masses appear clustered around stars of 0.7–1.2 M_☉, suggesting a stellar mass dependence in the evolution of the mass of millimeter-sized grains in disks.

Zoom In Zoom Out Reset image size

We calculate the Kendall's rank correlation statistic and find that the comparison is ambiguous when considering the entire sample. With a Z-value of 0.87, there is a 39% chance of rejecting the null-hypothesis that the disk mass and stellar mass are unrelated. Noting that the BA stars only host debris disks, we examine the stellar mass—dust mass relation for KM stars. In this case, the relationship is clear; with a Z-value of 3.17, there is only a 0.2% chance of disk dust mass not being related to the stellar mass. In Upper Sco, K stars tend to have disks with higher dust mass than do M stars. This could reflect numerous factors; these higher mass stars may have emerged from their natal clouds with higher mass disks or some mechanism may be active in this mass range to preserve disks.

The observation that millimeter-detected sources are not outliers in their NIR and Hα emission suggests that the process leading to low millimeter emission is not necessarily tied to the loss of disk gas nor small warm grains. However, the concentration of high dust masses around near-solar-mass stars suggests a stellar mass dependent mechanism in dust evolution. This expands upon the results of Carpenter et al. (2006) and Carpenter et al. (2009), who found evidence that small dust grains deplete more rapidly around stars of greater than solar mass than around stars of near-solar mass or less; our results suggest millimeter-emitting grains are retained longer around near-solar-mass stars.

The NIR and Hα classification systems we use above serve as indicators for the presence of dust and gas components of disks. While in younger groups the two properties are strongly correlated, in Upper Sco the majority of Class II systems show no evidence for accretion. The statistics of our observations and data collected from the literature indicate that we are seeing the tail-end of the primordial disk stage. All disks hosted by BA stars appear to be debris disks. Only 22 out of 127 (17%) KM stars show any evidence for remaining primordial disk material (either NIR excess, Hα emission, or mm-emission), and only 7 (6%) have disks detected in millimeter photometry.

These remaining primordial disks appear to be in the last stages of dissipation—their SEDs lie below the Taurus median, indicating significant inner disk dust evolution or flattening of the disk; they generally have low Hα equivalent widths, suggesting depleted gas reservoirs for sustained accretion; and their low mm fluxes indicate that the mass in millimeter-sized and smaller grains has, for the most part, decreased to the edge of detectability with current facilities.

One must keep in mind, however, that each of these measures reaches a different mass sensitivity. Because of this, one should exercise caution in interpreting the statistics regarding Class II/Class III sources and CTTS/WTTS sources.

Our millimeter observations reach a dust mass sensitivity of ∼4 × 10⁻³ M_Jup, while near-IR observations are sensitive to dust masses of ∼10⁻⁵ M_Jup (e.g., Cieza et al. 2007). These correspond to disk mass sensitivities of 0.4 M_Jup and 10⁻³ M_Jup, assuming the interstellar medium gas-to-dust ratio of 100 (e.g., Beckwith et al. 1990).

To estimate an equivalent disk mass sensitivity for Hα observations, we use the empirical relation of accretion rate as a function of time of Hartmann et al. (1998). Modeling the accretion rate ($\dot{M}$) as a power law in time ($\dot{M}\propto t^{-\eta }$, with stellar age t and η ≈ 1.5–2.8), one can estimate the remaining disk gas mass as $M_{{\rm gas}} \approx (0.5\hbox{--}2)\times \dot{M}(t)\times t$. The W_λ(Hα) cutoffs for classifying systems as accreting roughly correspond to accretion rates of 10⁻¹⁰ M_☉ yr⁻¹ (e.g., Dahm 2008). Thus, for our Upper Sco sources with t = 5 Myr, W_λ(Hα) observations provide an estimated disk mass sensitivity of ∼(0.3–1) × 10⁻³ M_☉. While this estimate spans a large range of values, and neglects many of the details of accretion (e.g., accretion variability, observation geometry), it illustrates that Hα observations are likely sensitive to disks of similar mass as those detected by our millimeter photometry. Both W_λ(Hα) and millimeter photometry based estimates of the disk mass are at least two orders of magnitude less sensitive than near-IR observations.

Therefore, one should expect the presence of numerous systems exhibiting infrared excesses, but lacking one or both of detectable Hα emission and detectable millimeter emission. This is indeed the case in Upper Sco, with only four sources exhibiting all three disk tracers, while 80% of accreting sources and all millimeter-detected sources exhibit an excess at 24 μm and shorter wavelengths.

HIP 76310. The sole mm-detected A star in our sample, HIP 76310, has the brightest infrared excess of the BA stars in Upper Sco. However, its infrared to stellar luminosity ratio (L_IR/L_* = 2.1 × 10⁻⁴, Dahm & Carpenter 2009) falls within the observational definition of a debris disk (L_IR/L_* < 10⁻³). Our assumption of 20 K for the average dust temperature of primordial disks is based on KM stars in Taurus and does not apply to the case of a debris disk around an A star. The SED fit by Dahm & Carpenter implies a much higher dust temperature of 122 K and thus a correspondingly lower dust mass (a factor of ∼8). Both our initial mass estimate (3.6 × 10⁻³ M_Jup) and revised dust mass of ∼5 × 10⁻⁴ M_Jup are several orders of magnitude higher than the lower limit of 2.6 × 10⁻⁷ M_Jup found by Dahm & Carpenter, highlighting the high NIR optical depth achieved by small quantities of dust, and the importance of (sub-)millimeter observations to directly constrain the dust mass. This dust mass also fits the evolutionary trend of debris disk dust masses as a function of time as shown in Williams & Andrews (2006).

${\textit [\,}\!{\it PZ}{\textit 99}\!\,{\textit ]} J160421.7-213028.$ The K2 star [PZ99] J160421.7-213028 stands out not only for having a uniquely steep infrared SED in Upper Sco, but also for having the largest millimeter emission. The SED shape for this object is typical of a transition disk, where dust grains have been cleared from the inner disk but are still abundant at larger radii. The strong millimeter emission implies a large dust mass (0.10 M_Jup). However, this source is not clearly accreting; the Hα equivalent width of −0.57 Å suggests that some process has greatly reduced or cut off accretion to the stellar surface. By current disk evolution models, the two most likely candidate processes are the onset of photoevaporation (Alexander et al. 2006) or the formation of a planet large enough to reduce stellar accretion below our detection limits (Alexander & Armitage 2007). Detailed examination of this disk requires millimeter imaging, which we defer to a later paper.

Binaries. ScoPMS 31 and [PZ99] J161411.0-230536 are detected in millimeter continuum and exhibit primordial type infrared excesses, yet have binary companions at projected separations of 84 and 32 AU, respectively (Köhler et al. 2000; Metchev & Hillenbrand 2009). There is ambiguity between observing a circumstellar or circumbinary disk; both cases allow for the presence of cold grains at large radii. While the question of the effect of binarity on disk mass and lifetime is worth considering, our numbers here are too small to provide a useful study.

At younger ages accreting systems dominate the population of sources with millimeter continuum detections. In Taurus, 67 out of 76 (88% ± 11%) millimeter-detected sources with Hα measurements in AW05 are accreting, while in IC348, all 9 mm-detected sources are accreting (LWC11). There is no such clear distinction in Upper Sco where, of the 7 millimeter detections, 4 are accreting and 3 are not.

Furthermore, as noted in Section 4.1, the Hα distributions of millimeter-detected and non-detected sources in Upper Sco are similar. This is a marked contrast to the younger IC348, where millimeter-detected disks are biased toward larger Hα equivalent widths than non-detected disks.

Examining the detection statistics in greater detail, we find that the disk detection rate among CTTS in Upper Sco is 44% ± 22%, significantly lower than that in Taurus. The detection rate for WTTS is similar. In Table 4, we show contingency tables constructed for detections and non-detections of CTTS and WTTS in the two regions. Using the Fisher exact test, we find for a two-tailed distribution the mm-detection rate among CTTS has ∼0.25% chance of being independent of region; the apparent drop in mm-detection rate for CTTS we see in Upper Sco is significant. For WTTS, the two-tailed Fisher's exact test indicates that the mm-detection rate for WTTS remains the same between Taurus and Upper Sco. While the mm-detection rate for WTTS has changed little from 1 to 5 Myr, the detection rate for CTTS has decreased markedly.

The millimeter-detection rate for Class II sources drops significantly from 1 to 5 Myr. In Taurus, 86% of Class II sources have millimeter detections (AW05), while in Upper Sco that fraction has dropped to 32% ± 13% (6/19). Furthermore, the median disk mass decreases. Taurus Class II disks have a median dust mass of 3 × 10⁻² M_Jup, while in Upper Sco the median dust mass of detected disks is 1.3 × 10⁻² M_Jup. The true median is likely to be lower due to the high censoring rate in our data.

The Class III sources in Upper Sco are unlikely to host additional massive disks. We did not detect any of the nine Class III disks in our survey, and the inferred upper limit to the detection rate of <11% is consistent with the detection rate for Class III sources in AW05, 7% ± 4%.

As with the accretion state, we construct contingency tables of the detection statistics of Taurus and Upper Sco Class II and Class III sources. We show the number of detections of each type of object in Taurus and Upper Sco in Table 4. As with CTTS, we find that the Class II sources are detected at a lower rate in Upper Sco than in Taurus. The detection rate for Class III sources, on the other hand, is not clearly distinguishable.

Due to the variation in exposure times in our observations, and hence sensitivities, we use the Kaplan–Meier product limit estimator for randomly censored data sets (Feigelson & Nelson 1985) to generate the cumulative dust mass distribution functions, F(⩾ M_Dust), for Class II sources in these regions (Figure 5). The dust mass distribution of Upper Sco Class II sources is clearly different from that of Taurus Class II sources, with the upper quartile of Upper Sco sources overlapping the lowest quartile of Taurus sources; 27% ± 10% of Upper Sco sources have disks with a dust mass greater than or equal to 4 × 10⁻³ M_Jup, compared to 89% ± 4% of those in Taurus. We make use of the Gehan, Peto-Prentice, and log-rank statistics to compare the highly censored dust mass distributions of Upper Sco and IC 348; the distributions are different, with mild statistical significance (there is only a 2.5%–7% chance that the distributions are drawn from the same parent function).

Zoom In Zoom Out Reset image size

Having established that the dust mass distribution has changed from 1 to 5 Myr, we examine the rate at which the highest dust masses change. Comparing similarly constructed mass distributions, LWC11 find that the fraction of IC348 disks with M_dust greater than 2 × 10⁻² M_Jup is approximately equal to the fraction of Taurus disks with M_dust greater than 40 × 10⁻² M_Jup, suggesting a systematic factor of 20 decrease in the dust mass of Class II sources from 1 Myr to 2.5 Myr. If this trend were to continue, one would not expect to find any disks with millimeter continuum emission in Upper Sco. However, our results indicate that dust masses do not decline so greatly from 2.5 to 5 Myr; the fraction of disks with M_dust > 2 × 10⁻² M_Jup is similar between IC348 and Upper Sco, and the highest masses for Class II disks in Upper Sco are lower than masses of IC348 disks by only a factor of ≲ 2.

These results indicate that for Class II sources, the observed dust mass rapidly declines from 1 to 2.5 Myr, but may only slowly change from then until 5 Myr. In contrast, the fraction of disks exhibiting IR-excess consistent with primordial dust continues its marked decline. In Taurus, 50% of KM stars are Class I or Class II (Luhman et al. 2010). We include Class I sources here as their disks consist of primordial material. In IC348, only 39% of KM stars are Class II, with 1% of sources being Class I (Spitzer photometry of Lada et al. 2006; Muench et al. 2007). In Upper Sco the Class II fraction is 15%, with no Class I sources.

While there is no clear evidence for a large age spread among Upper Sco stars, Preibisch et al. (2002) could not rule out a 1–2 Myr age spread. This allows for the possibility of some overlap in stellar ages with the stars of IC 348 (where there may be an age spread of several Myr). It is possible that the mm-detected sources in Upper Sco are also the youngest sources. If these sources were only marginally older than the sources observed in IC 348, the small decrease in the maximum mm-emission for Class II disks would be consistent with a continuation of the steep drop in mm-emission seen in comparing Taurus to IC 348.

However, this argument could be equally applied to IC 348 and Taurus, with the corresponding claim that only the youngest disks are detected in the survey of LWC11, and the most massive disks in Taurus are also the youngest. Assuming similar age spreads in these regions, this scenario would not change the general picture we find of rapid change in the greatest disk masses over several Myr, followed by a more gradual decline.

There have been recent attempts to trace the evolution of disk properties as a function of stellar age (e.g., Isella et al. 2009; Guilloteau et al. 2011). The size of star forming regions, however, suggest distance uncertainties to individual stars large enough to cause 30% or greater uncertainties in luminosity estimates; other factors (e.g., photometric errors) can make this uncertainty even higher. In turn, 30% luminosity uncertainties are large enough to cause age uncertainties of several Myr (Hillenbrand et al. 2008). Until the age of individual stars can be accurately determined, interpretation may need to be restricted to consideration of average ages.

Comparison of HR diagrams for these regions indicates that the median age in Upper Sco is older than Taurus and IC 348 (see Figure 1 of Hillenbrand et al. 2008). The comparison between Taurus and IC 348 is more ambiguous; across all temperatures, the median sequences could be equivalent within the luminosity errors. However, at temperatures of log T_eff less than 3.625 (corresponding to a spectral type of about K6), the median Taurus sequence has luminosity equal to or higher than the IC 348 sequence, suggesting that the median Taurus age for KM stars is likely younger than IC 348. These interpretations of the HR diagrams are consistent with the decrease in Class II and CTTS fraction seen from Taurus to IC 348, and then further to Upper Sco. While the uncertainty in absolute ages of star forming regions, age spread within those regions, and ages of individual stars complicate interpretation, the overall trend is clear.

Significant progress has been made in recent years in modeling the dust evolution in disks. The amount of millimeter sized grains will decrease over time as dust aggregates to form centimeter and larger bodies, to which millimeter photometry is insensitive (Dullemond & Dominik 2005). Birnstiel et al. (2010b) modeled mm-slopes and fluxes in disks where millimeter-size grains remained present due to an equilibrium state between coagulation and fragmentation. Their models are consistent with observations of young disks at 0.5–2 Myr, and later work (Birnstiel et al. 2011) shows the same mechanisms can maintain populations of micron-sized grains that produce NIR excess emission. Observations of millimeter continuum emission from older disks, such as this work, provide important constraints on the later time evolution of dust, and perhaps constrain the mechanisms leading to loss of observable dust (e.g., radial drift, agglomeration into centimeter and larger grains).

In summary, there appears to be a rapid evolution of disks from 1 to 2.5 Myr, an interval during which both millimeter-dust masses and disk fractions rapidly drop. While the infrared disk fraction continues to decrease from 2.5 to 5 Myr, the dust masses of the remaining disks do not rapidly change. In addition, disks appear to reach a low baseline of dust mass where there is no longer a correlation between dust and inner disk gas.

In addition to providing a point of comparison for the evolution of circumstellar disk mass as a function of age and inner disk properties, our disk dust mass measurements provide new insights on giant planet formation.

Under the assumption of a primordial 100:1 gas-to-dust mass ratio (e.g., Beckwith et al. 1990) as if planet formation has yet to begin, Upper Sco retains only a single known Minimum-Mass Solar Nebula disk (MMSN, 10 M_Jup, Weidenschilling 1977), the 10 M_Jup disk around [PZ99] J160421.7-213028. If planet formation were to begin now in Upper Sco, then the region could not form Jupiter-mass planets at the frequency observed in the field (∼10%, Johnson 2009). However, the disks in Upper Sco are clearly evolved, and the dust has grown substantially. Jupiter-mass planets must have already formed, or be well on their way to formation, as seen in the younger sources of IC348 (LWC11). The discovery of planetary companions to the Upper Sco members 1RXS J160929.1-210524 and GSC 06214-00210 (Lafrenière et al. 2008; Ireland et al. 2011) serve as confirmation to this scenario. Due to our greater sensitivity, however, we can push this argument further to examine the formation of Neptune-mass planets.

At the 5 Myr age of Upper Sco, large rock and ice protoplanets may have formed, or may be in the process of formation. There are still uncertainties in the growth of grains from centimeter to planetesimal (10⁵ m) sizes, but simulations indicate growth from micron to centimeter size grains occurs in as little as 2 × 10⁵ years (Birnstiel et al. 2010a), and planetesimals can aggregate to form protoplanets with enough mass to accrete gas (∼0.03 M_Jup) in less than 1 Myr (Pollack et al. 1996). Under core accretion models, the formation of a gas giant planet requires that gas be retained in the disk until the accreting protoplanet reaches the critical mass (∼0.1 M_Jup) at which runaway accretion begins; this process could take from less than 1 Myr (Alibert et al. 2005) to several tens of Myr (Pollack et al. 1996), depending on several factors such as disk surface density, formation radius, and the effects of planetary migration.

Debris disks are expected to be seen in systems in the final stages of giant planet formation, as well as after the conclusion of giant planet formation. Our results verify that BA stars in Upper Sco have no residual cold millimeter-size dust above debris levels; these disks are unlikely to retain substantial gas mass, signaling the end of giant planet formation.

However, some lower mass KM stars have NIR or mm continuum emission and/or Hα emission indicating the presence of gas and dust. The median inferred mass for the handful of mm-detected disks, under the assumption of the primordial 100:1 gas-to-dust ratio, is ∼0.1 MMSN. By linear scaling of the MMSN, such a disk could produce a planet approximately twice the mass of Neptune (M_N = 0.05 M_Jup) if planet formation were to begin now. Only four sources have disks with inferred masses of 0.1 MMSN or greater, implying ∼2% of stars in the 5 Myr Upper Sco group retain disks able to form a 2M_N planet. In comparison, results of the first 4.5 months of the Kepler mission suggest that up to 19% of field stars have Neptune-size planet candidates within 0.5 AU (Borucki et al. 2011). The formation of Neptune-mass planets, in addition to the formation of Jupiter-mass planets, must be at an advanced stage by 5 Myr.