This thesis present new results on the nature of brown dwarf atmospheres revealed by the analysis based on the spectra taken by the infrared astronomical satellite AKARI and a brown dwarf atmosphere model. The main results of this thesis are summarized as follows. (1) We construct a spectral data set of brown dwarfs that continuously covers a new wavelength range, 2.5-5.0 μm, (2) we verify a generality of the molecular abundances deviating from a thermo-chemical equilibrium state for T dwarfs, (3) we show that L5 dwarfs with or without the CH4 3.3 μm band in their spectra differ in dust presence and mass (surface gravity) in addition to Teff, (4) we verify the theoretical prediction of radius inversion observationally for the first time, (5) we argue that the dust contribution to the atmospheric structure is larger than that in the current models, and (6) we suggest a possibility that relative C and 0 elemental abundances with respect to H are different in every objects.

Brown dwarfs are objects that are too light to maintain the hydrogen fusion in their cores. Their effective temperatures are very low as 2200-600 K. They are classified into newly introduced spectral types L and T. L dwarfs are warmer than T dwarfs. The study of brown dwarfs has been actively carried out since 1995, when the first genuine brown dwarf, G1 229B, was discovered by Nakajima et al. Brown dwarfs are important as a bridge, between stars and planets. With intermediate masses and temperatures, brown dwarfs are expected to have the blended properties of star and planet. However, their properties are actually quite unique, for example dusty photosphere, and it is not straightforward to understand their internal physical and chemical processes from our knowledges of stars and planets. Studies of brown dwarf atmosphere will lead us to understand the comprehensive nature of "atmospheres" of various objects from stars to planets.

Atmospheres of brown dwarfs are so cool and are dominated by molecules. When temperature decreases below the condensation temperature, dust forms in the photosphere. Dust affects the spectra of brown dwarfs by changing the photospheric structure and by extinction. The effects of dust are observed most apparently in the near-infrared spectra of L dwarfs. On the other hand, the spectra of T dwarfs show little sign of dust. This indicates that the dust in the photosphere disappears somewhere between L and T types. Since the mechanism of dust disappearance is not yet clear, current brown dwarf photosphere models implement dust segregation effect empirically through a model parameter. Such models can explain the observed SED more or less satisfactory. On the other hand, the models still have problems to explain some molecular absorption bands especially in the spectra of late-L to T dwarfs. It is not clear whether the deviation of molecular abundance from theoretical prediction is general characteristics or not. This inconsistency is one of the essential problems in the study of brown dwarf atmospheres.

Spectroscopic observations in the infrared regime are the most powerful tools to obtain physical and chemical information of brown dwarf atmospheres through various molecular bands. So far, brown dwarf spectra shorter than 2.5 μm have been obtained by ground-based observations. The brown dwarf atmospheres have been investigated with the data, but it is difficult to discuss molecular abundances in detail and systematically because of weak and blended molecular overtone absorption bands. The wavelength range between 2.5 and 5.0 μm is the most suitable for this purpose as it contains features of major molecular spices, including the CH4 v3 fundamental band at 3.3 μm, CO2 v3 fundamental band at 4.2 μm, CO fundamental band at 4.6 μm and H2O vi and v3 absorption bands around 2.7 μm. Since these fundamental bands of important molecules are non-blended each other, we can investigate these molecular bands in detail. However, observations in the wavelength range from the ground is always challenging. Severe absorption due to the Earth's atmosphere and limited wavelength coverage make the precise analysis difficult.

The Japanese infrared astronomical satellite AKARI was launched in February 2006. The InfraRed Camera (IRC) on-board AKARI is capable of yielding moderate-resolution (R~120) spectra in. this important wavelength range devoid of any degradation by telluric features. We have conducted an observing program using the IRC to obtain continuous spectra of brown dwarfs in 2.5-5.0 μm wavelengths aiming to carry out systematic studies of physical and chemical processes in their atmospheres.

Twenty seven brown dwarfs, sixteen L dwarfs and eleven T dwarfs, were successfully observed by AKARI. The standard software toolkit IRC SPEC TOOLKIT was used for the data reduction. Wavelength and flux calibrations were all done in the toolkit. Spectral data are derived from two dimensional spectral image. We applied the following three additional processing to obtain better quality data; (1) derivation of appropriate sky background, (2) stacking of multiple observation data, and (3) correction of contaminated light from nearby objects. As a result, we obtain 18 continuous spectra of brown dwarfs from 2.5 to 5.0 μm for the first time (Figure 1). We validate the absolute flux calibration of the obtained AKARI spectra by comparing the integrated flux measured in the spectra with (i) the AKARI photometry data obtained simultaneously during the observation, and (ii) past photometry data in literatures. We find that absolute flux of the spectra is consistent with the both AKARl photometry and past photometry data within 15% and 10%, respectively.

As a first step, we investigate the behavior of various non-blended molecular fundamental bands in brown dwarf spectra, the CH4 at 3.3 μm, CO at 4.6 μm and CO2 at 4.2 μm, relative to their spectral types systematically with the 18 AKARI observed spectra. Carbon resides mostly in CH4 rather than in CO in very cool (e.g. T < 1000K) and high density environment. The temperature of brown dwarf atmospheres is just around the boundary that CH4 molecule starts appearing. The behavior of the CH4 3.3 pm fundamental band has not been investigated in detail because of fragmented and small number of data samples. It is also difficult to observe the CO 4.6 pm fundamental band. In the past study, one observed data of T8 dwarf shows that the CO absorption band strength is stronger than theoretical prediction, but its generality has not been ensured. We find that the CH4 3.3 μm fundamental band appears in the spectra of dwarfs later than L5 and CO 4.6 μm band appears in the spectra of all spectral types until late-T dwarfs (Figure!). Investigation of the CO2 absorption band is the first attempt for the brown dwarf atmosphere. We detected CO2 absorption band at 4.2 μm in the spectra of late-L and T type dwarfs. The result for the non-blended CO molecular band is very important because of the fact that CO generally exists in all brown dwarf atmospheres, which is against theoretical predictions. We also find that the CO2 molecule is generally in the atmosphere of T dwarfs.

To understand the atmospheres of brown dwarfs better, we analyze the AKARI spectra using the Unified Cloudy Model (UCM) developed by Tsuji et al., which is one of the brown dwarf atmosphere models. We derive the physical parameters, effective temperature Teff, surface gravity log g and critical temperature Ter, of AKARI samples by model fitting. We investigate how the parameters correlate with the spectral type. We confirm that the spectral types of late-L dwarfs are not a sequence of Teff, but a sequence of the decreasing of dust effect.

We investigate the property of the objects in which the CH4 3.3 μm band starts to appear. We find that the band is seen in two of four L5 dwarfs. We evaluate the physical condition of the photosphere of the objects by applying the UCM. We find that the model parameters of the sources with and without the CH4 3.3 μm band are systematically different, except for the effective temperature. Surface gravity and critical temperature, which is an additional parameter to determine the upper limit of the dust layer, of the objects with the CH4 3.3 μm band are higher than those of the objects without the band. Our model fitting analysis confirms that the appearance of the CH4 absorption band at 3.3 μm in L5 type spectra depends on not only their effective temperatures but also dust presences and surface gravities. We suggest that the two groups are different in their masses.

Theoretical models of brown dwarf evolution predict that the radius of a young object follows a monotonically increasing function of mass and a decreasing function of age. On the other hand, for old object (〓108 yrs) the dependency of radius on mass inverts, i.e. the radii of less Theoretical models of brown dwarf evolution predict that the radius of a young object follows a monotonically increasing function of mass and a decreasing function of age. On the other hand, for old object (>=10(8) yrs) the dependency of radius on mass inverts, i.e. the radii of less Figure 2 Radius of AKARI observed brown dwarfs. The decreasing radius between early- to late-L dwarfs and the increasing radius to late-T dwarfs may indicate that radius inversion predicted by theories is detected by current analysis. massive objects become larger because the less massive objects reach the terminal radii early. We derive the radii of the 18 AKARI objects using their parallaxes and the ratio of observed to model fluxes, and observationally verify this radius changes for the first time (Figure 2).

In this thesis we use the archived near-infrared spectra (IRTF/SpeX and UKIRT/CGS4, hereafter SpeX/CGS4) covering the shorter wavelength range (1.0-2.5 μm) along with AKARI to derive the physical parameters of the AKARI objects. With the additional short wavelength data we can constraint the model fitting better. However, a concern is that we could not determine a unique model that explains the entire wavelength range perfectly, and there are always some deviation. This has been pointed out in previous studies. In order to investigate this problem further, we fit the AKARI and SpeXICGS4 spectra separately. We find that the AKARI spectra is more sensitive to the effective temperature, while the dust presence is better determined by the SpeXICGS4 data. When the SpeX/CGS4 spectra show the presence of large amount of dust, the effective temperatures derived only from AKARI spectra show a higher value than those determined using the A.KARI +SpeXICGS4 data. The AKARI spectra also show higher effective temperatures when the SpeX/ CGS4 spectra show a small amount of dust. These results imply that the warming up effect due to the dust in the photosphere is always underestimated in the model than that in the actual photosphere. Dust opacity is sensitive to the grain size distribution, their amount and composition. We propose that a self-consistent, more realistic theory of condensation and sedimentation in the atmospheres is the most essential in the future brown dwarf atmosphere models.

The observed CO2 absorption band in some objects is stronger or weaker than the prediction by the model. We discuss possible metallicity variation among brown dwarfs using the model atmosphere and the AKARI data. We construct a set of models with various elemental abundances as a first trial. We investigate the variation of the molecular composition and atmospheric structure. From the results, we suggest a possible reason of CO2 4.2 μm absorption feature in the late-L and T type spectra is the higher or lower C and 0 elemental abundances than the solar values used in the previous studies.

Figure 1 AKARI spectral data of 19 brown dwarfs observed for the first time. The position of CH4, CO2 and CO fundamental bands are shown in yellow, green and blue respectively.

Figure 2 Radius of AKARI observed brown dwarfs. The decreasing radius between early- to late-L dwarfs and the increasing radius to late-T dwarfs may indicate that radius inversion predicted by theories is detected by current analysis.

本論文は、赤外線天文衛星「あかり」のデータによる褐色矮星の研究である。近赤外線(波長1-5μm)は、褐色矮星のスペクトルがピークとなる重要な波長帯である。地上観測が中心だったこれまでは、主として1-2.5μmのデータに基づいた研究が行われてきた。しかし、この波長では星の大気中に存在する主要分子のバンドが重なって、スペクトルの解析が困難である。一方2.5-5μmでは、主要分子の基底バンドはスペクトル上で分離して存在するが、地上観測では良質のデータが得られない。本研究は、広いスペクトル型の範囲にわたる褐色矮星18個について、宇宙空間からの観測の特徴を生かして2.5-5μmのスペクトルを取得・解析し、褐色矮星の大気構造を研究したものである。

論文の第1章では、褐色矮星の定義、研究の歴史、スペクトル型の分類と特徴、未解決の問題などが、専門外の研究者でも理解できるように丁寧に解説されている。

第2章では、「あかり」衛星と観測装置の説明があり、続いて観測とデータ処理について解説されている。

第3章は、観測から直接導くことができる結果の記述である。本研究では、11個のL型星と7個のT型星について、2.5-5μmにわたる連続した良質のスペクトルを得ることに初めて成功した。続いて、スペクトルに見られる吸収バンド(H2O 2.5μmと4.7μmより長波長、CO 4.6μm、 CO2 4.2μm、CH4 3.3μm)が説明され、定量化のため各種指数が導入される。COはすべてのスペクトル型で検出されているが、特にT型晩期星からの検出は熱化学平衡モデルでは説明できないことが言及される。

第4章では、論文提出者が得た「あかり」衛星による2.5-5μmのデータに、これまで地上観測で得られていた1-2.5μmのデータを付加して、統一ダスト雲モデル(UCM)を適用した結果が記述されている。スペクトルフィットにおいては、有効温度(Teff)、表面重力加速度(log g)、ダストが消失する臨界温度(Tcr)を自由パラメータとし、他に元素組成などは太陽組成を仮定するなどした。フィットの結果、L型晩期(L6-T0)の星では有効温度に100 K以上の変化は見られず、スペクトル型の違いは臨界温度の違いを反映している可能性があることを指摘した。また、観測した4個のL5型星において、2個からはCH4が検出され、残りでは検出されていないが、この違いが表面重力の違い、すなわち質量の違いによる可能性があることを示唆した。さらに、求めた有効温度とフラックスから半径を計算し、T型褐色矮星の半径が、晩期になるに連れて木星の0.7倍から1.1倍程度まで増加することを見出した。これは、質量が小さいほど半径が大きいことを意味し、縮退圧で支えられた星で理論的に予想される性質と定性的に合致する。

第4章では、「あかり」衛星のデータ(2.5-5μm)と地上観測(1-2.5μm)のデータを、UCMを用いて同時にフィットするには困難があることが述べられるが、第5章はその原因について考察がなされている。そのため、2.5-5μmのデータと、1-2.5μmのデータに独立にUCMを適用し、2.5-5μmのスペクトルは有効温度に敏感であるのに対し、1-2.5μmのスペクトルは褐色矮星大気中のダスト吸収に敏感であることが帰結される。そこから、1-5μmにわたるスペクトルをモデルで統一的に説明するためには、現在のモデルよりダストの影響を大きくする必要があることを指摘している。

第6章では、前章までフィッティングの対象から除外してきた波長4.2μmにあるCO2の吸収バンドの強さを説明することを試みている。そのために、前章までのモデルでは太陽組成に固定してきた元素組成を、特にC、O、Feに注目して変化させている。その結果、Feを太陽組成としてCとOの組成比を変化させることで、CO2の吸収バンドもフィットできることを示している。

以上の結果は、「あかり」衛星の特徴を最大限に生かした研究によって、以下の6点について重要かつ新たな知見を与えたものである。(1) スペクトル型L1-T8の広い有効温度範囲(2,200-600 K)にある18個の褐色矮星について、波長2.5μmから5μmまでの完全なスペクトルを初めて取得したこと。(2) 大部分の炭素原子がCH4として存在すると考えられていたT型晩期星の大気からも明確にCOを検出し、T型褐色矮星大気中の分子組成が熱化学平衡からずれていることを示したこと。(3) 4個のL5型星で見られたCH4の吸収バンドの有無が、これらの褐色矮星の質量の違いに起因する可能性を指摘したこと。(4) 古い褐色矮星は質量が大きいほど半径が小さいという、縮退星に特有の傾向を初めて観測的に示したこと。(5) 現状の褐色矮星大気モデルでは、2.5-5μmのスペクトルと、これまで得られていた1-2.5μmのスペクトルを矛盾なく説明することは困難であり、この矛盾を解決するひとつの可能性として、大気中のダストの影響をより大きくする必要があることを指摘したこと。(6) 褐色矮星大気中のCO2の検出に初めて成功し、その吸収強度を理解するためには、CとOの組成が太陽組成と異なる必要があると示したこと。

なお、本論文は辻隆氏、山村一誠氏との共同研究であるが、論文提出者が主体となって観測、解析、考察に重要な役割を果たしており、論文提出者の寄与が十分であると判断する。

したがって、博士(理学)の学位を授与できると認める。