A&A 418, 117-129 (2004)
N. Babkovskaia - J. Poutanen
Astronomy Division, PO Box 3000, 90014 University of Oulu, Finland
Received 15 September 2003 / Accepted 22 December 2003
We study in detail a pumping mechanism for the cm maser transition in ortho-H2O based on the difference between gas and dust temperatures. The upper maser level is populated radiatively through and transitions. The heat sink is realized by absorbing the 45 photons, corresponding to the transition, by cold dust. We compute the inversion of maser level populations in the optically thick medium as a function of the hydrogen concentration, the water-to-dust mass ratio, and the difference between the gas and the dust temperatures. The main results of the numerical simulations are interpreted in terms of a simplified four-level model. We show that the maser strength depends mostly on the product of hydrogen concentration and the dust-to-water mass ratio but not on the size distribution of the dust particles or their type. We also suggest approximate formulae that describe accurately the inversion and can be used for fast calculations of the maser luminosity. Depending on the gas temperature, the maximum maser luminosity is reached when the water concentration times the dust-to-hydrogen mass ratio, and the inversion completely disappears at densities just an order of magnitude larger. For a dust temperature of 130 K, the transition becomes inverted already at a temperature difference of K, while other possible masing transitions require a larger K. We identify the region of the parameter space where other ortho- and para-water masing transitions can appear.
Key words: ISM: dust, extinction - masers - radio lines: general - methods: analytical - methods: numerical
Sources of strong water maser emission at wavelength cm have been found in many astrophysical objects such as active galactic nuclei, carbon-rich stars, protostellar regions, and comets. Maser emission is a powerful tool for investigating the physical conditions in the emitting regions because of its high brightness as well as high sensitivity to the physical parameters of the medium in which the maser amplification takes place.
There are several mechanisms for the ortho-H2O ( cm) maser pumping. De Jong (1973) proposed a comprehensive model where the breakdown of thermal equilibrium occurs in the surface layers of the dense gas cloud. Upon approach to the surface, the transition (45 ) (which is among the most important for the maser action, see Fig. 1) becomes transparent first, and level 523 becomes underpopulated. The de Jong model was widely used to describe the water masers from late-type stars and star-forming regions (Cooke & Elitzur 1985; Elitzur et al. 1989; Neufeld & Melnick 1991), and the molecular accretion disks in active galactic nuclei (Babkovskaya & Varshalovich 2000; Neufeld et al. 1994).
The de Jong mechanism is, however, not able to explain the most powerful masers in star-forming regions. A different physical process has to be involved. In shocks, the departures from equilibrium could be produced by collisions with hotter "superthermal'' hydrogen gas (Varshalovich et al. 1983). Strelnitskij (1984,1980) proposed that collisions with two species (e.g., charged and neutral particles) of different temperatures and with comparable collision rate can produce the necessary inversion at high hydrogen densities needed to explain the high observed maser luminosities. Kylafis & Norman (1987) pointed out that the magnetohydrodynamic shocks naturally produce conditions where the electron temperature is larger than the hydrogen temperature . In such a case, the lower maser level becomes underpopulated, because the relative importance of collisions with neutrals is larger for the transitions to/from this level. The inversion in this model appeared only in a narrow range of the ionization fraction 10-5-10-4 and for K. At high ionization, electrons start to dominate the collisions, and the levels are thermalized at , while at lower ionization neutrals are dominating and the thermalization happens at . Using better estimates for the collision rates, Elitzur & Fuqua (1989) and Anderson & Watson (1990) ruled out this model unless the neutral particles are hotter than the charged ones, conditions that are difficult to imagine in any astrophysical environment.
All astrophysical objects that show water maser emission are also expected to have non-negligible quantities of dust. If the dust and the gas have different temperatures, departures from equilibrium are possible (Kegel 1975; Goldreich & Kwan 1974; Strelnitskij 1977; Bolgova et al. 1977). Goldreich & Kwan (1974) suggested that the radiation from the hot dust excites the water molecules to the vibrational state, and the heat sink is provided by collisions with cooler (than dust) hydrogen molecules. The possibility of inversion in such a situation was questioned by Deguchi (1981) who pointed out that because the collisional de-excitation rate between vibrational states is much smaller than the pure-rotational collisional rate, the rotational collisional thermalization may quench the maser when vibrational collisional de-excitation becomes dominant (see also Strelnitskij 1988).
Alternatively, the cold dust can produce the necessary inversion (Strelnitskij 1977; Bolgova et al. 1977). In an optically thick environment, the excitation temperature takes a value between the dust and the gas temperatures depending on the relative role of dust and collisions in the destruction of the line photons. Deguchi (1981) considered the following cycle of maser pumping (see Fig. 1). He showed that the downward transition at 45 is much more affected by dust absorption than the upward transitions at , because there is a strong peak near 45 in the absorption coefficient of the cosmic-type ices (e.g., Moore & Hudson 1994). The upper level excitation temperature is then close to the gas temperature, while the lower level becomes populated at the dust temperature. Similar inversion occurs for other types of dust too.
One should note that in this model an arbitrarily thick layer can participate in the maser action provided the gas and dust temperatures differ sufficiently. Chandra et al. (1984a) computed the maser efficiency when both the surface escape and the cold dust absorption mechanisms operate together. Recently Collison & Watson (1995) and Wallin & Watson (1997) applied this model to the masing disk in the active galaxy NGC 4258, concluding that the cold dust model is much more efficient (see also Neufeld 2000). Thus the cold dust-hot gas model seems to be the most promising to explain powerful water masers in many astrophysical objects. The temperature difference between the gas and the dust can appear as a result of shock heating or illumination by the UV- or X-ray photons, and/or because of the presence of the dust particles of different types and sizes which assure different temperatures.
The purpose of the present study is to determine the maser strength for a broad range of physical parameters: hydrogen concentration, water-to-dust mass ratio, gas and dust temperatures as well as dust type. Previous studies (Chandra et al. 1984a; Deguchi 1981) used a rectangular instead of a Doppler line profile as well as the collisional coefficients with hydrogen of Green (1980), which later have been much improved (Green et al. 1993). Only a handful of parameters was explored in the more recent investigation of Collison & Watson (1995) and Wallin & Watson (1997) who considered the case of saturated masers.
In its full statement this problem needs the simultaneous solution of the statistical balance equations together with the radiative transfer equations for all spectral lines. However, because the cold dust - hot gas model can operate inside a molecular cloud where almost all transitions (except masing) are optically thick, the radiative transfer can be handled in a very simple manner using the escape probability formalism, where now the role of the escape of spectral line photons is played by the dust absorption.
The paper is constructed as follows. We give the formulation of the problem and describe the numerical method in Sect. 2. We develop a simple four-level model that contains most of the physics involved in Sect. 3. The numerical results for different dust types and size distributions are presented in Sect. 4, where we also propose simple analytical formulae for the inversion which describe the results of simulations. Finally, in Sect. 5 we present our conclusions.
|Figure 1: A portion of the diagram of ortho-H2O showing the rotational levels involved in the pumping of the maser.|
|Open with DEXTER|
We consider a slab consisting of a mixture of molecular hydrogen,
water molecules and dust particles.
The radiative transfer equation for the specific intensity Ixdescribing the transfer of the line radiation
in the presence of continuum absorption and emission is (Mihalas 1978; Hummer 1968)
The line-averaged absorption coefficient
(including induced emission)
depends on the corresponding level populations:
Let us denote the optical thickness of the slab in ul-line
averaged over the line profile as
A formal solution of the radiative transfer
for a homogeneous (i.e.
slab gives the mean intensity of the radiation averaged
over the line profile at the optical depth
(Mihalas 1978; Hummer 1968):
If the source function does not vary much, we can
take it out from the integral (the so called on-the-spot or the first order
escape probability approximation) and get:
Following Hollenbach & McKee (1979), we approximate
The masing lines need a different treatment.
We neglect the dust influence in such transitions, i.e. assume
Using expression (10), one can show that at
large maser optical depth
We consider different types of dust. The amorphous ice absorption coefficient is computed from the optical constants of Leger et al. (1983). The crystalline ice absorption coefficients are based on the data from Bertie et al. (1969) and Warren (1984). The silicate and graphite data are from Laor & Draine (1993).
The populations of M levels can be determined from the
population balance equations which, in the stationary case, take the form:
Substituting the radiation intensity from Eq. (9) into
Eq. (21) and the statistical balance Eq. (19),
The populations can be found from the solution of the system of non-linear algebraic Eqs. (12), (18), and (22). We use the Newton-Raphson method with line searches and backtracking (Press et al. 1992). We normally start from high hydrogen density where the solution is close to the Boltzmann distribution at the gas temperature and use the obtained solution as the zeroth approximation for the next point. We continue iterations until the maximum error in the system becomes smaller than 10-11 and the maximal change in populations is smaller than 10-9. It takes 5 to 15 iterations for most sets of parameters and each computation takes on average about 15 ms CPU time on a Pentium IV 2 GHz Linux PC.
If the maser amplification takes place in the homogeneous
medium and there is no velocity gradient,
the maser intensity can be estimated as:
The water level populations depend on the hydrogen
the dust and water concentrations and
temperatures as well as the slab half-thickness H.
When the photon escape from the surface is negligible because
the slab is optically thick either in continuum (
in the line (), the terms K2 and L2 can be
omitted in the escape probabilities (12) (i.e.
The solution of the population balance equations then depends
which in turn depends only on the ratio of
the absorption coefficients ,
on the absorption coefficients individually. Since
the dust opacity
has a rather
weak dependence on the size of the dust particles, it
is rather natural to use
the water-to-dust mass ratio
as a parameter,
is the water-to-hydrogen mass ratio and
is the dust-to-hydrogen mass ratio.
The water concentration is then related to these parameters
(the factor 9 comes from the
When the escape through the surface is small and the masers are unsaturated
(i.e. a masing transition does not affect the level populations),
the populations do not depend on
the scale-height H, and the optical depth is then
linearly proportional to
In such a situation the main parameters are
, T and .
If the dust is sufficiently cold, its own
radiation is negligible and dependence on
We normally consider the dust colder than the gas,
because this is required for inversion to occur (in the optically thick case).
|Figure 2: Dependence of the 616-523maser absorption coefficient on the half-thickness of the slab for six sets of parameters : (1) 109,10-2; (2) 1010,10-3; (3) 1011,10-4; (4) 109,10-4; (5) 1010,10-5; (6) 1011,10-6. Other parameters are fixed at K, T=250 K, . At a slab thickness less that 1013 cm, the effect of the boundary becomes significant. For higher H, the maser becomes saturated when .|
|Open with DEXTER|
One can also point out that for large when in the main infrared pumping lines, the solution of the balance Eq. (22) depends on the product , but not on and individually, because .
We consider a range of hydrogen concentrations from 108 to . At higher , the inversion is absent because of the thermalization by collisions with the hydrogen molecules, while at smaller we can extrapolate the results from .
In astrophysical objects it is easier to estimate the dust-to-gas
than other parameters of interest. We
fix it at a value standard for the interstellar medium of
This then determines the dust optical depth
The water-to-dust mass ratio probably varies by orders of magnitude from one object to another. Therefore, we consider a broad range of from 10-8 to 100. The maximum possible in the interstellar medium is about 10-3 which transforms to the maximum for . However, for a smaller dust content, can be larger.
|Figure 3: Levels of constant maser absorption coefficient (solid curves) and maser optical depth (dotted curves) in the plane slab half-thickness - gas temperature for and . Other parameters are fixed at K, .|
|Open with DEXTER|
Figure 2 shows the absolute value of the absorption coefficient
(in units of
in the main masing line
for six representative
sets of parameters
One sees that at
cm, the surface
escape of photons (the de Jong mechanism) begins to influence
At high H, the optical depth
can be sufficiently large to saturate the maser.
One can easily estimate the optical depth when it happens.
The saturation occurs when the induced transition rate
in the masing line becomes comparable to the rates of transitions
from the upper masing level in other strong lines
(see Eq. (21)) which are
Because the Einstein A coefficient in the 1.35 cm
masing line is about 10-9(and about 10-5 in other masing transitions), the saturation
On the other hand, since
The inversion is rather flat around cm, where it is a function of the local conditions only. Therefore, we use this height for our calculations. We should also note that our results do not depend on the geometry of the system, and can be applied not only to the slab but to any other geometry.
The contours of constant maser absorption coefficient and the optical depth in a plane H-gas temperature Tare shown in Fig. 3. One sees that, independently of the gas temperature, the absorption coefficient is rather flat around cm. The flat part becomes shorter at higher gas temperature, because the saturation ( ) happens for smaller H. One should note here that the parameter set taken for this graph gives much shorter flat parts than other sets presented in Fig. 2. Our fiducial H=1014 cm still seems to be a good choice for studying in detail the physics of the unsaturated water maser in a dusty environment. It is also interesting to note that for small H inversion exist even when because of the action of the de Jong mechanism.
Our main goal is to study in detail the inversion mechanism
Before we proceed to the results of the calculations, it is useful
to understand the physical processes responsible for the action of this main
maser using a simplified system of molecular levels which mainly participate
in the maser pumping: 414 (level 1), 505 (level 2),
523 (level 3) and 616 (level 4) (see Fig. 1 and
The upper maser level 4 interacts
mostly with the 2nd level which in turn interacts with level 1,
while the lower maser level 3 interacts directly mostly with level 1.
(We neglect here rather strong transitions from the 2nd to the 3rd level via
The rates of radiative (in the case of the unsaturated maser)
as well as collisional transitions via the masing line
are much smaller than the rate of transitions to other levels.
Thus, we can assume that the populations at the masing levels
are completely unrelated to each other, and depend only on the
rate of exchange to other levels.
This allows us to write a system of three (two-level) population balance equations
that relate the corresponding populations in the standard form:
When the water concentration is high, the source functions depend on one parameter . This means that the inversion is the function of only. If for all main lines is very small (e.g. small hydrogen density) then the levels are thermalized at the dust temperature, while for large , thermalization occurs at the gas temperature. The limits on where the inversion is possible depend on .
Let us first investigate the limit .
The source functions can then be represented as
In another extreme case, small influence of the dust relative to collisions,
Now keeping only the term
we obtain a condition for the inversion
We can get an estimate for the temperature difference
needed to produce the inversion.
The inversion appears first in the region where the
is dominated by dust (i.e.
while for other main transitions the dust influence
is still negligible (i.e.
in the vicinity of
we get from the condition
A low water content (or high dust content) is described by
For a slab thickness H=1014 cm,
the medium is transparent for small
(since the dust optical depth is proportional to the
hydrogen density), and thus a similar analytical
description to the one above is not valid.
However, for larger H, and/or
the medium still can be sufficiently opaque.
Then, the source functions in all pumping lines are
Analogously, we can get an upper limit on
keeping the terms of the first order in
Now let us obtain the minimum temperature difference needed for the inversion.
Representing the source function as
for small ,
Let us compare the de Jong model to the cold dust model.
Using the escape probability approximation,
the radiation field is now
where p is the probability for photons to escape without interactions and
we neglected here the influence by the dust. The source function is then
|Figure 4: Contour plots of the levels of constant inversion (dotted curves) and constant optical depth (solid curves) for the 616-523 transition in the plane hydrogen concentration - water-to-dust mass ratio. A slab of half-thickness H=1014 cm and dust in the form of amorphous ice of the size was assumed, with a dust-to-gas mass ratio . The panels correspond to the following temperatures a) K, T=250 K; b) K, T=132 K; c) K, T=250 K; d) K, T=400 K. The dashed contour in panel b) corresponds to the negligible photon escape case .|
|Open with DEXTER|
|Figure 5: Contour plots of the levels of constant 616-523 maser optical depth (solid curves) in the plane gas-dust temperature difference - water-to-dust mass ratio for the following set of parameters , K, , and H=1014 cm. The dashed curves show the corresponding contours if the de Jong mechanism is not operating (i.e. the photon escape from the slab is neglected and the K2 and L2 terms in Eq. (12) are omitted).|
|Open with DEXTER|
The main results of our calculations for the case of amorphous ice of the size of are shown in Figs. 4-6. The dependence of the inversion and of the maser optical depth on and for fixed dust and gas temperatures is shown in Fig. 4. One sees that when the temperature difference is small (Fig. 4b), the inversion appears in the region where within two orders of magnitude. The inversion is weak for small . We tested the influence of the de Jong mechanism on the results in this region by neglecting the terms K2 and L2 in the escape probabilities (12). In that case (dashed curve in Fig. 4b), the inversion is absent there because a higher temperature difference is needed (see Eq. (43)) to invert the populations.
When is larger (see Figs. 4c, d), the inversion becomes larger and the inversion region increases in size, also spilling over into the region with a low water content ( ). Here the radiation field is completely dominated by the dust and the inversion depends on the collisional rates defined by only (for the fixed T). At lower dust temperature K (Fig. 4a), the optical depth becomes larger by about 20% compared to K (Fig. 4c).
We should point out that the inversion disappears at because of the level thermalization by collisions. This is lower than our analytical estimate (42) which neglected many collisional transitions. On the right side the inversion region is bounded by for T=250 K, which is very close to our estimate (37).
Let us note that the maximum inversion occurs when the water-to-dust mass ratio . The maser optical depth, on the other hand, is proportional to which reaches the maximum at for T=250 K. This corresponds to . The inversion disappears at because the amount of dust (comparing with water) is not sufficient here to produce the inversion. Thus for a given dust content there exists an optimal water concentration that produces the strongest maser. In order to calculate the maser intensity (see Eq. (23)) or the brightness temperature in the line center , one needs a rough estimate for the excitation temperature. Because the dependence is linear, the error in is not important. At the maximum of , the excitation temperature varies between -10 and -100 K when T varies between 150 and 500 K.
|Figure 6: Temperature dependence of the maser optical depth for the same six sets of parameters as in Fig. 2. We fixed H=1014 cm, K, . The solid curve corresponds to analytical formula (45) with free normalization.|
|Open with DEXTER|
In Fig. 5 we present the dependence of on and for a fixed value of . We see that the inversion disappears at large because the levels are then thermalized at the gas temperature. At small , the medium is optically thin for the line and the dust emission ( for most infrared lines at this for ). Here we also tested the influence of photon escape from the surface. The dashed curves show the maser optical depth when the escape is neglected. These results show that at small the inversion is produced by the de Jong mechanism, while without it the minimum to invert the population is about K. For small , however, the inversion appears only for K. At higher this limit is 3.5 K (see dashed curve in Fig. 9) which is very close to the predictions of our four-level model (43).
Figure 6 shows the dependence of the optical depth in the main masing transition on the gas temperature for the same six sets of parameters as in Fig. 2. One sees that for most parameter sets the inversion appears when gas temperature exceeds . For set 4, however, the inversion exists even at smaller T. This results from the action of the de Jong mechanism in an optically thin (in line) medium with a low dust content (see also Fig. 3).
The optical depth increases sharply at small
saturates at K. This behavior is accurately
described by our analytical formula (32).
The optical depth is thus proportional to
Table 1: Ortho-water masing transitions.
Table 2: Para-water masing transitions.
Our simulations show that many other masers appear
in addition to the
The strongest masers for ortho-water are listed in Table 1 and for para-water
in Table 2.
The levels of constant optical depth are
shown in Figs. 7 and 8, respectively.
Some of these masers were found in the calculations of
Chandra et al. (1984a) and Wallin & Watson (1997).
Most of them appear also in models based on the de Jong (1973)
mechanism (see also Cooke & Elitzur 1985; Neufeld & Melnick 1991).
been detected in the Orion cloud by Phillips et al. (1980) and Waters et al. (1980).
The masing emission in
line was observed
in a wide variety of sources where the 1.35 cm maser was detected
(Menten et al. 1990).
|Figure 7: Contours of constant maser optical depth for the strongest masing transitions of ortho-water for K, T=400 K. The other parameters are the same as in Fig. 4.|
|Open with DEXTER|
|Figure 9: Isolines of the zero inversion of the strongest masing transitions for the same parameters as in Fig. 5. The labels on the contours correspond to the numbering in Table 1. Panel a) is for ; b) corresponds to . The inversion of the maser level populations exists on the right side of the corresponding isoline. The dashed curves on the left show the zero inversion for the maser when the de Jong mechanism is turned off.|
|Open with DEXTER|
|Figure 10: Same as Fig. 9, but for para-water. The labels on the contours correspond to the numbering in Table 2.|
|Open with DEXTER|
We show the isolines of the zero inversion of the level populations as a function of the gas temperature for fixed in Figs. 9 and 10. The inversion of the level populations appears on the right side of the corresponding curves. We see that the inversion in most of the transitions arises when the difference between the gas and dust temperature is larger than 30-100 K, and one does not expect to see a significant signal in other water masing transitions if is small. This is related to the fact that the energy separation between the corresponding levels is much larger than for the maser. Many masers also require the water concentration to be two orders of magnitude smaller than that at which the and masers are strongest (see Figs. 7 and 8). In these respects the cold dust - hot gas model significantly differs from the de Jong model where many masing transitions appear simultaneously.
Since the optical properties of the ice in the far-infrared spectral range for different grain sizes are very similar, the maser strength is almost independent of a. The crystalline ice has a stronger peak in at and therefore the 616-523 maser optical depth is 30% larger than for the amorphous ice. We repeated calculations with silicate and graphite grains. The optical depth decreases by 15% and 25%, respectively.
We also computed the maser conditions for a mixture of the graphite and SiC dust. We consider the size distribution of the dust grains , which extends from to . To allow easy comparison with the previous results (Collison & Watson 1995; Wallin & Watson 1997), we use an approximation of the results by Laor & Draine (1993, see their Fig. 6) for the dust cross-section. We assume a dust opacity in the form for and , for . The resulting maser optical depth is very similar to that for the crystalline ice.
The maser luminosity is a function of the optical depth which depends on the inversion . In many astrophysical problems, one would like to make simple estimations of the inversion and optical depth, not repeating cumbersome calculations of the water molecular level populations. Thus one would like to have simple analytical approximate formulae for this purpose.
It is possible to design a formula for the inversion that describes it with an error of a factor of two using the four-level model from Sect. 3. We can further simplify the model by assuming that the upper masing level 616 is populated directly from the 414 level. We then arrive at a system of two equations, each corresponding to the two-level model. We now can introduce a pseudo-transition , and ascribe to it some collisional and radiative rates.
We propose the following formula to compute the inversion:
We consider the cold dust - hot gas mechanism of pumping of the water masers. To obtain the inversion of the maser level populations we take into account the first 45 rotational levels of ortho- and para-H2O molecules and all possible collisional and radiative transitions between them. In the case of large optical depth in the main pumping lines and unsaturated masers, the radiative transfer problem can be significantly simplified, allowing us to investigate the maser efficiency for a set of the main parameters, such as hydrogen concentration, water-to-dust mass ratio and gas temperature.
As was suggested by Deguchi (1981), the pumping cycle mainly determines the inversion of the 616-523 maser levels populations. We use this four-level model for the analysis of our numerical results. We also suggest approximate formulae for the dependence of the inversion on the hydrogen concentration, the gas-to-dust mass ratio and the gas and dust temperatures, that could be used for modeling the astrophysical sources. We find that the inversion is largest when the water-to-dust mass ratio is , and the maser optical depth is the largest when .
The inversion of the maser level populations as a function of the same parameters is calculated for amorphous and crystalline ice, silicate, and graphite dust. The results of these calculations are very similar. Because the inversion depends on the dust mass density, our results can be applied to any size distribution of the dust particles.
Our analysis shows that in optically thick environments the inversion in the transition appears when the gas is just 1 K hotter than the dust, while most other transitions begin to be inverted at much larger K. Most of them require also a smaller water-to-dust mass ratio . This is very different from the predictions of the de Jong model where many transitions are inverted simultaneously. If the cold dust - hot gas model is the correct one, the relative ratios of the luminosities in different masing transitions could provide constraints on the gas and dust temperatures, water-to-dust mass ratio, and hydrogen concentration.
This work was supported by the Centre for International Mobility, the Magnus Ehrnrooth Foundation, the Finnish Graduate School for Astronomy and Space Physics, and the Academy of Finland. We are grateful to Ryszard Szczerba for providing the dust absorption coefficients, to Dmitrii Nagirner for providing the codes for computations of K- and L-functions, and to Vsevolod Ivanov for useful comments on the manuscript and discussions.