New measurements on water ice photodesorption and product formation under ultraviolet irradiation

Abstract

The photodesorption of icy grain mantles has been claimed to be responsible for the abundance of gas-phase molecules towards cold regions. Being water a ubiquitous molecule, it is crucial to understand its role in photochemistry and its behaviour under an ultraviolet field. We report new measurements on the ultraviolet (UV) photodesorption of water ice and its H₂, OH, and O₂ photoproducts using a calibrated quadrupole mass spectrometer. Solid water was deposited under ultra-high-vacuum conditions and then UV-irradiated at various temperatures starting from 8 K with a microwave discharged hydrogen lamp. Deuterated water was used for confirmation of the results. We found a photodesorption yield of 1.3 × 10⁻³ molecules per incident photon for water and 0.7 × 10⁻³ molecules per incident photon for deuterated water at the lowest irradiation temperature, 8 K. The photodesorption yield per absorbed photon is given and comparison with astrophysical scenarios, where water ice photodesorption could account for the presence of gas-phase water towards cold regions in the absence of a thermal desorption process, is addressed.

methods: laboratory: molecular, protoplanetary discs, ISM: molecules, ultraviolet: ISM

1 INTRODUCTION

Water is everywhere. This statement has been proven by different space missions like the Infrared Space Observatory (ISO). Gas phase water emission has been found in contrasting astrophysical scenarios like shocks, photodissociation regions (PDRs), and high and low-mass protostars (e.g. Cernicharo & Crovisier 2005). On the other hand, water ice was reported to exist in several ISO sources such as embedded young stellar objects (YSOs) and in molecular clouds towards the line of sight of field stars. This offered a characterization of ice composition in a variety of lines of sights (e.g. Gibb et al. 2004). Herschel Space Observatory observed thousands of YSOs (Pilbratt et al. 2010). Using PACS (Photodetector Array Camera and Spectrometer), several groups detected water emission in objects with different stellar associations, sources, and spatial extensions (Meeus et al. 2010; Tilling et al. 2012; Howard et al. 2013; Thi et al. 2013; Lindberg et al. 2014; Riviere-Marichalar et al. 2015, 2016; and references therein). Using ALMA (Atacama Large Millimeter/submillimeter Array), Cieza et al. (2016) detected indirectly the water snow line of the object V883 Ori, concluding that snow lines might have highly dynamical behaviours and they should be considered when a disc evolution model is under development.

Water ice controls the efficacy of dust particles and planetesimals aggregation by enhancing their stickiness (see Blum & Wurm 2008 and references therein). Water photodesorption plays an important role in protoplanetary discs. Making use of computational models, Akinori, Taishi & Shigeru (2012) reported that photodesorption holds down the ice formation sweeping the snow line notably outwards. In addition, the snow line position can be highly determined by the equilibrium between photodesorption and ice formation. Not taking into account the photodesorption effect in models translates into a deeper ice absorption feature than astronomical observations. In the meantime, models with photodesorption show a shallow absorption feature nevertheless, and the later seems to be a better match (see Honda et al. 2016).

From the laboratory point of view, water ice has been the subject of study for many years. Greenberg et al. (1980) showed one of the first experiments at low temperature (10 K) under high-vacuum conditions (10⁻⁸ mbar). They deposited an ice mantle made of water, carbon monoxide, ammonia, and carbon dioxide. After ultraviolet (UV) irradiation, new molecules were produced as a result. Years later, Allamandola, Sandford & Valero (1988) started performing water matrix experiments to re-create a more realistic interstellar ice mantle analogue. Westley et al (1995a,b) reported the first measurements on water photodesorption by ultraviolet irradiation. Muñoz Caro & Schutte (2003) and Öberg et al. (2010) investigated the effects of water concentration on ice mantles and its importance on photochemistry. They found that in water matrix studies the destruction of the molecules under ultraviolet irradiation increases with water concentration. This has a direct effect on the molecule lifetime under protostellar conditions.

In this paper, we report laboratory experiments of pure water ice under UV irradiation and ultra-high-vacuum (UHV) conditions. Deuterated water was used to confirm the data. Here we present measurements of the photodesorption yields of water, deuterated water, and their photoproducts. We provide quadrupole mass spectrometry calibrated data during the UV irradiation and our results are compared with previous experimental and theoretical works. The paper is organized as follows. The methods are explained in Section 2. We present our results in Section 3, followed by a discussion in Section 4, and the conclusions are addressed in Section 5.

2 METHODS

The ISAC set-up was used to conduct laboratory experiments of solid water photoprocessing (see Muñoz Caro et al. 2010). It has a base pressure typically in the range P = (3.0–4.0) × 10⁻¹¹ mbar and a minimal temperature of 8 K, making use of a closed-cycle helium cryostat, which allows us to grow ice layers by deposition of gas species on a cold infrared (IR) transparent window. The solid samples are processed by UV radiation and they can be monitored with in situ FTIR spectroscopy in transmittance, while molecules desorbing to the gas phase are detected by a quadrupole mass spectrometer (QMS).

For the experimental simulations here described, we used H₂O (liquid), triple distilled, and D₂O (liquid), Cambridge Isotope Laboratories, Inc (C.I.L.) 99.9 per cent. In our experiments, the deposition of water vapor is directed using a capillary outlet towards the substrate window; this outlet is separated about 2 cm from the substrate. Therefore, most of the water is condensed on the substrate; for more details, see Muñoz Caro et al. (2010). The ice mantles were irradiated using a microwave discharged hydrogen flow lamp (MDHL) which has a strong Ly-α and low-molecular-hydrogen emission simulating the radiation experienced by interstellar ice mantles. The UV flux is ≈2 × 10¹⁴ photons cm⁻² s⁻¹ with a mean energy of 8.6 eV (strong Ly-α centred at 10.2 eV and molecular hydrogen emissions around 7.8 eV) at the sample position (Muñoz Caro et al. 2010; Chen et al. 2014; Cruz-Diaz et al. 2014a). The water (and deuterated water) ice samples were deposited at 8 K and subsequently warmed up to the irradiation temperature of 30, 50, 70, and 90 K. After this, the samples were irradiated with cumulative intervals that led to total irradiation times of 1, 3, 10, 30, and 60 min. FTIR spectroscopy was performed after deposition of the ice samples, and between each irradiation dose, to monitor the evolution of the original ice component. In particular, the column densities of solid water and solid deuterium oxide were calculated from their IR absorption using the formula: |$N = \frac{1}{\mathcal {A}} \int _{{\rm band}} \tau _{\nu } {\rm d}\nu$|⁠, where N is the column density in molecules per cm², |$\mathcal {A}$| the band strength in cm molecule⁻¹, τ_ν the optical depth of the band, and dν the wavenumber differential in cm⁻¹. The adopted band strengths were (1.7 ± 0.2) × 10⁻¹⁶ for the 3259 cm⁻¹ band of H₂O (Cruz-Diaz et al. 2014a) and (1.0 ± 0.2) × 10⁻¹⁶ for the 2413 cm⁻¹ band of D₂O (Cruz-Diaz, Muñoz Caro & Chen 2014b). Table 1 summarizes the ice thickness for water and deuterated water experiments, which are only slightly higher than the canonical interstellar ice thicknesses (Zubko, Dwek & Arendt 2004).

Table 1.

Water and deuterated water irradiation experiments performed for this study.

H2O	Irrad. temp.	Thickness
sample No.	(K)	(ML)
1	8	203
2	30	210
3	50	205
4	70	201
5	90	211
D2O
Sample No.
6	8	245
7	30	251
8	50	248
9	70	249
10	90	252

Notes. One monolayer (ML) corresponds to 1 × 10¹⁵ molecules cm⁻².

The number photons absorbed by the ice thickness can be derived using the Beer-Lambert's law,

\begin{eqnarray} I_{\rm t} = I_0 \, {\rm exp}(-\sigma N) \end{eqnarray}

(1)

where I_t is the transmitted intensity, I₀ the incident intensity, N is the ice column density in cm⁻², and σ is the cross-section in cm². N was taken from Table 1, I₀ is the photon flux, and σ the average VUV absorption cross-section of solid H₂O or D₂O: 3.4 × 10⁻¹⁸ cm² and 2.7 × 10⁻¹⁸ cm², respectively (see Cruz-Diaz, Muñoz Caro, & Chen 2014a,b, ). We found that on average for the water experiments 50  per cent of the photons were absorbed while on average for deuterated water was 51  per cent.

The main photoproducts during water irradiation, according to Westley et al. (1995a,b), Gerakines, Schutte & Ehrenfreund (1996), and Öberg et al (2009), are H₂O₂, HO₂, O₂, OH and H₂. H₂ and O₂ do not present IR features but we see them using a QMS. No IR features corresponding to the remaining photoproducts were detected during these experimental simulations. This is an effect of the low conversion rate from water to hydrogen peroxide, ∼0.1  per cent. This would be translated into a production of ∼2ML of H₂O₂, close to the detection limit. We cannot measure the amount of hydrogen or oxygen trapped in the ice bulk, but since the former is very volatile and the ladder is not produced by a primary photodissociation process of H₂O, we assume that their concentrations are low and the ice sample can be treated as an almost pure water ice. Therefore, we do not find relevant to show the IR spectra collected before and after the UV photoprocessing because we want to focus on the gas data.

As mentioned above, we used quadrupole mass spectrometry (QMS) to detect the gas-phase molecules that desorbed thermally or by an UV photon from the ice mantle during the irradiation period. QMS measurements were carefully calibrated using the method described in Martín-Doménech et al. (2015), allowing conversion from the integrated ion current of a particular mass fragment A(m/z) to photodesorbed column density N(mol cm⁻²). This method uses additional pure CO ice irradiation experiments to derive a proportionality constant k_CO between the integrated ion current measured for the m/z = 28 mass fragment of photodesorbing CO molecules and the decrease of the CO ice column density measured with the FTIR spectrometer during these additional experiments, which is mainly due to photodesorption.

\begin{eqnarray} k_{{\rm CO}} = \frac{A(28)}{N(CO)}. \end{eqnarray}

(2)

This proportionality constant k_CO can be used to quantify photodesorption of other species like H₂O monitored with a particular mass fragment m/z by applying the different factors in the equation:

\begin{eqnarray} N({\rm mol}) &=& \frac{A(m/z)}{k_{{\rm CO}}} \times \frac{\sigma ({\rm CO})}{\sigma ({\rm mol})} \times \frac{{\rm IF}({\rm CO}^+)}{{\rm IF}(z)} \nonumber \\ &&\times\,\frac{{\rm FF}(28)}{{\rm FF}(m)} \times \frac{S(28)}{S(m/z)} \end{eqnarray}

(3)

with σ the ionization cross-section in the mass spectrometer for the first ionization of CO and the species of interest, knowing that the electron energy for ionization in the QMS is 70 eV, IF the fraction of ionized molecules with the particular charge of the monitored mass fragment, FF the fraction of molecules of the isotopologue of interest leading to the monitored fragment in the mass spectrometer, and S the mass sensitivity of the QMS, which is probed regularly using noble gases. For more details on the calibration process, we refer to Martín-Doménech et al. (2015). For calculation purposes, we took: |$\sigma _{{\rm H}_{\rm 2}{\rm O}}$| = 2.097 × 10⁻¹⁶ cm², |$\sigma _{{\rm D}_{\rm 2}{\rm O}}$| = 2.049 × 10⁻¹⁶ cm² (Straub et al. 1998), |$\sigma _{{\rm D}_{\rm 2}}$| = 1.02 × 10⁻¹⁶ cm² (Rapp & Englander-Golden 1965, tables 77 and 161 in Celiberto et al. 2001, and Yoon et al. 2010), σ_OH = 1.85 × 10⁻¹⁶ cm² (Deutsch, Becker & Märk 1997), and |$\sigma _{{\rm O}_{\rm 2}}$| = 2.44 × 10⁻¹⁶ cm² (Rapp & Englander-Golden 1965).

3 EXPERIMENTAL RESULTS

Fig. 1 shows the mass spectrum of H₂O and D₂O measured using our QMS apparatus during the deposition of the ice sample. The main mass fragment is 18 for water and 20 for deuterated water, followed by 17 and 16, and 18 and 16, respectively. H₂O and D₂O main IR features are centred around 3290 and 2470 cm⁻¹. Although we performed IR spectroscopy to monitor the deposited ice mantle, we measured the photodesorption yield using our QMS because the main band of water and deuterated water are affected by the dissociation of the molecule, in other words, production of OH and OD radicals. In particular, the OH absorption feature overlaps with the broad H₂O absorption band around 3290 cm⁻¹ (O–H stretching mode), and the O₂ product is a homonuclear diatomic molecule with no dipole which makes it inactive in the IR.

Figure 1.

Mass spectrum of water (red bars) and deuterated water (black bars) obtained during deposition of the ice sample using our QMS apparatus at 8 K. The spectra are normalized relative to the main mass fragment (m/z = 18 for H₂O and m/z = 20 for D₂O).

3.1 Experiments at 8 k

Experiments were performed at different temperatures as stated previously, but the most relevant temperature to look for a pure photodesorption effect is 8 K. At this temperature the only volatile molecules free enough to diffuse through the ice bulk and subsequently desorb thermally are H₂ and D₂. The influence of heavier molecules diffusing throughout the ice bulk could be considered as low since thermal desorption of other detected molecules, O₂ and H₂O, occurs at higher temperatures, near 30  and 160 K, respectively (e.g. Collins et al. 2004). Fig. 2 shows the photodesorption observed at 8 K for H₂O/D₂O, O₂, and H₂/D₂ during the different irradiation intervals. We were able to measure the photodesorption yield of water and deuterated water deposited at 8 K for different temperatures of irradiation. Table 2 shows the average measured photodesorption yields at different irradiation temperatures for H₂O, D₂O, O₂, OH, and D₂.

Figure 2.

QMS data measuring the photodesorption of different species by UV irradiation at 8 K. Exp. 1 denotes the irradiation of a pure H₂O ice mantle at 8 K, while Exp. 2 denotes the irradiation of a pure D₂O ice mantle at 8 K.

Table 2.

Photodesorption yield of water, deuterated water, and its photoproducts at different irradiation temperatures. Data from different groups have been added for comparison.

Irrad. Temp.	This work				Westley et al.	Andersson	Öberg et al.	Arasa et al.	Arasa et al.
					(1995a,b)	& van Dishoeck (2008)	(2009)	(2010)	(2011)
Kelvin	H2	OH	H2O	O2	H2O
	×10−3|$\left[\displaystyle\frac{{\rm mol}.}{{\rm phot.}}\right]$|
8	*	0.7 ± 0.3	1.3 ± 0.2	0.6 ± 0.1
10						0.14–0.37		0.54
20								0.57
30	*	0.5 ± 0.2	1.4 ± 0.4	1.1 ± 0.3				0.71
35					3.46
40
50	*	0.6 ± 0.3	1.6 ± 0.3	1.4 ± 0.2	3.63
60
70	*	1.0 ± 0.4	2.4 ± 0.6	2.4 ± 0.5	4.48
85					5.61
90	*	1.1 ± 0.5	2.5 ± 0.7	3.6 ± 0.9				0.71
100					7.63
Irrad. Temp.	D2	OD	D2O	O2		D2O
Kelvin	× 10−3|$\left[\displaystyle\frac{{\rm mol}.}{{\rm phot}.}\right]$|
8	2.7 ± 0.7	*	0.7 ± 0.1	0.5 ± 0.2
10									1.27
20							2.07		1.46
30	11.4 ± 2.2	*	0.9 ± 0.2	1.2 ± 0.2			2.33
35
40							3.22
50	14.3 ± 2.3	*	1.0 ± 0.2	1.3 ± 0.3
60							3.50		1.57
70	18.5 ± 3.4	*	1.6 ± 0.4	2.5 ± 0.5
85
90	15.7 ± 3.1	*	1.5 ± 0.3	3.3 ± 0.8					1.83
100							4.27

Notes. The photodesorption yield measured is the average number of molecules desorbed per incident photon for the total irradiation time at selected temperature. *These values suffer from background H₂ or H₂O contamination in the chamber. The latter has the same m/z value of 18 as OD. Therefore, we only provide the values for D₂ and OH, respectively.

Because of fluctuations in the QMS signal (see Figs 2 and 3 top), we cannot say the H₂O/D₂O photodesorption is constant. However, we can appreciate an average photodeorption for the entire experiment where each irradiation step is within the expected error (see Fig. 3 bottom). Comparing water and oxygen behaviours in this figure, we consider the photodesorption of water and deuterated water at different irradiation periods to be UV-fluence independent. This is not the case for oxygen where the photodesorption yield clearly grows with irradiation time. Typically an error of about 12  per cent in the photodesorption yields was estimated as the result of UV-flux oscillations during the irradiation of the ice and the noise superposed on the QMS signals (see grey area in Fig. 3 top). Although one of the main contaminants inside the chamber is water, QMS data were calibrated and water from the background was removed (see the Appendix section for a full description of the method).

Figure 3.

Top: Close-up of Fig. 2 showing the noise in the QMS and the small bumps caused by instabilities in the H₂ flux used for the UV lamp. Exp. 1 denotes the irradiation of a pure H₂O ice mantle at 8 K. Bottom: Photodesorption yield as a function of fluence for water (dots), deuterated water (asterisks) and oxygen (diamonds) at 8 K. H₂O and D₂O present no dependence with the fluence. Solid and dotted lines represent the average photodesorption yields for water and deuterated water in Table 2.

The OH/OD intensities in the QMS are lower than that of the parent molecule, which leads to a lower signal-to-noise ratio. In the case of OD, its m/z = 18 coincides with the molecular mass fragment of the background H₂O. The resolution of our QMS does not allow us to distinguish between these two mass fragments; for this reason we did not provide the value of its photodesorption yield in Table 2.

Pure CO ice presents a constant photodesorption rate for ice thickness above 5 ML (Muñoz Caro et al. 2010; Fayolle et al. 2011; Chen et al. 2014). The lower photodesorption rate of H₂O did not allow the determination of a critical ice thickness. In the case of photoproducts that desorbed either thermally or during irradiation from the ice mantle we detected H₂/D₂, OH/OD, and O₂.

3.2 H₂O/D₂O photodesorption

We performed UV irradiation of H₂O/D₂O at different temperatures to study its impact on photodesorption. A previous article reported a similar study using methanol and deuterated methanol (see Cruz-Diaz et al. 2016). We found CH₃OH and CD₃OD photodesorption was too low to detect using our QMS, providing a photodesorption upper limit of 3 × 10⁻⁵ molecules per incident photon, in line with Bertin et al. (2016). This is not the case for water ice. Fig. 4 shows the QMS signal for water and deuterated water (using their respective main mass fragments) during the ice irradiation at different temperatures.

Figure 4.

Photodesorption of H₂O and D₂O deposited at 8 K during UV irradiation at different temperatures. The different steps follow the irradiation (step) and non-irradiation (flat signal) of the sample by the UV lamp. This plot follows the colour code in Fig. 2. Black = 8 K, blue = 30 K, green = 50 K, yellow = 70 K, and red = 90 K.

We observed the typical step-like features that correspond to the various irradiation intervals, indicating a photodesorption event from the ice mantle at every irradiation temperature. Notice that for temperatures of 30 K and above the photodesorption of water becomes UV-fluence dependent, see Fig. 5 to appreciate better the behaviour. This could be a consequence of O₂ thermal desorption enhancing the photodesorption yield of H₂O.

Figure 5.

Photodesorption yield at different irradiation periods for temperatures of 30 K and above for water (left) and deuterated water (right).

Zhu et al. (1993) summoned the MGR model to explain the difference in photodesorption between NH₃ and ND₃. They found that the photodesorption of NH₃ is four times stronger than the one for ND₃. It was concluded that a mass effect in the N–H coordinate was responsible for the isotopic effect in NH₃ photodesorption. This model predicts an isotope effect stating that a lighter particle is accelerated in a shorter time on the excited potential energy surface. Therefore, this particle has a higher probability of escaping from the substrate. This effect has been seen in surface photon-driven processes. Adapting equation (2) in their publication for our experiments, we obtain:

\begin{eqnarray} \frac{\sigma _d({\rm H}_{2}{\rm O})}{\sigma _d({\rm D}_{2}{\rm O})} = \left(\frac{1}{P_d({\rm H}_{2}{\rm O})}\right)^{\sqrt{\frac{m_{D}}{m_{H}}}-1} \end{eqnarray}

(4)

where σ_d(H₂O) and σ_d(D₂O) are the photodesorption cross-sections, in our experiment at 8 K |$\frac{\sigma _d({\rm H}_{2}{\rm O})}{\sigma _d({\rm D}_{2}{\rm O})}$| is about 2.21 (see Table 2), |$\frac{m_{{\rm D}}}{m_{{\rm H}}} = 2$| for the masses of D and H atoms is an approximation of the reduced masses in the O–D and O–H stretching coordinates. This leads to a photodesorption probability of P_d(H₂O) = 0.146, i.e. nearly 15  per cent of the photon-excited molecules is desorbing. The MGR model could explain the difference we observe between H₂O and D₂O signal intensities in the QMS after compensation from background H₂O contamination. H₂O photodesorption is two times stronger than the one for D₂O.

3.3 H₂/D₂, OH/OD, and O₂ photodesorption

Hydrogen can diffuse through the ice sample at any temperature in our experiments. It is also the main contaminant inside the chamber as it is common in UHV chambers with stainless steel walls, contributing to the large difference between H₂ (m/z = 2) and D₂ (m/z = 4) QMS signals (H₂ signal almost doubles D₂) in Fig. 6. This could be also an isotope effect as in the case of H₂O and D₂O. It can be observed that the intensity of the QMS signal is UV-fluence independent at 8 K, suggesting that H₂/D₂ are formed efficiently and desorb very readily at a constant photon-induced desorption rate. H₂ can form by diffusion and reaction of H produced by photodissociation of H₂O giving H + OH, or directly from H₂O photodissociation into H₂ + O (e.g. Okabe 1978). We also observed the increase in the QMS signal with temperature, similar to water desorption data. In the case of H₂/D₂ a larger diffusion at higher temperatures can enhance the desorption up to almost six times higher. Since H₂ suffers from background contamination, only the D₂ photodesorption yield is provided in Table 2. D₂ photodesorption at higher temperatures than 8 K presents a particular behaviour (see Fig. 6). D₂ signal presents a maximum photodesorption at the first irradiation period of 1 min, while it slowly decreases for each consequent irradiation. We are not sure about what is causing this behaviour. Molecular hydrogen production by UV irradiation using pure methanol or pure water ice is remarkably different. The H₂ photodesorption in water is three orders of magnitude lower in this work compared to Cruz-Diaz et al. (2016) for methanol.

Figure 6.

Photodesorption of H₂ (top) and D₂ (bottom) during irradiation of water and deuterated water ice at the different irradiation doses and temperatures. The difference in QMS signal intensity is at least partially due to H₂ background contamination. This plot follows the colour code in Fig. 4.

During H₂O/D₂O ice irradiation, OH (m/z = 17) and OD (m/z = 18) photodesorption is observed besides their low intensity in the QMS signal (data were smoothed for a better appreciation of the step-like features) (see Fig. 7). Only the OH photodesorption yield is given in Table 2 since OD signal could be contaminated by background water since they share the m/z = 18. We took into account the portion of H₂ and OH signals coming from the dissociation of H₂O inside the QMS apparatus using the ratio in Fig. 1. Because the fragmentation ratio of H₂O on the filament of the QMS may vary between experiments (this is an empirical result; it is a behaviour we have noticed during the time the QMS have been active. We cannot explain it but we can monitor the issue and know when it happens therefore we can correct), the values provided for D₂ and OH have a significant error. OH photodesorption is UV-fluence independent during irradiation at 8 K.

Figure 7.

Photodesorption of OH (top) and OD (bottom) during irradiation of water and deuterated water ice at the different irradiation doses and temperatures. The data have been smoothed to help in the visualization of the photodesorption events. This plot follows the colour code in Fig. 4.

Molecular oxygen, O₂, has a strong QMS signal during its photodesorption (see Fig. 8). It reaches the same intensity in both experiments (H₂O and D₂O irradiation), which increases progressively with irradiation time up to a maximum, denoting a bulk and a fluence dependence effect. This explains the fluence dependence of molecular oxygen photodesorption plotted in Fig 3 bottom (see Cruz-Diaz et al. 2016 for a detailed description of this effect). This trend in the photodesorption of O₂ is expected because O₂ is not formed by a primary photodissociation process of H₂O. At high ice temperatures, diffusion of O atoms is enhanced and O₂ photodesorption increases. It is worth to notice that the kinetic observed with the light on/off intervals follows that of a continuous irradiation over the explored fluence range. We observed that the end of an irradiation coincides with the start of the other in the QMS signal intensity (see Fig. 8 for a clear example).

Figure 8.

Photodesorption of O₂ at different irradiation doses and temperatures. It presents the step-like function progressively increasing with irradiation time, denoting a bulk effect and a UV-fluence dependence. (*) Oxygen QMS signal during D₂O irradiation. This plot follows the colour code in Fig. 4.

Atomic oxygen, H₂O₂, and O₃ were not found to desorb upon UV irradiation in our experiments. We found that the main contribution to m/z =16 of O⁺ was due to H₂O/D₂O and O₂ fractionation in the QMS filament. Therefore, based on our data, it was not possible to infer the photodesorption of oxygen atoms. We found no evidence for H₂O₂ or O₃ formation. The former was not seen in IR or QMS data (m/z = 34) which indicates that its abundance is too low. Ozone (m/z = 48) IR features were not observed neither its QMS signal. This could be because, if produced, it would be dissociated into O₂ + O.

4 DISCUSSION

Many attempts to measure the photodesorption of water have been published over the past years. It is such a hot topic mainly because of the implications of water photodesorption in astronomical research. This work is justified by the current disagreement between experimental values from different teams. Our method, after calibration of our QMS, can give the total amount of photodesorbing molecules directly measured in the gas phase, and not indirectly in the solid phase. Here we compare our results to those of selected works previously published, all of them indirectly measuring the photodesorption yield.

OH has been detected towards Sgr B2 with an abundance of 2 × 10⁻⁶ to 5 × 10⁻⁶ (Goicoechea & Cernicharo, 2002). They attributed this OH abundance to a photon-dominated region. Our experiments give proof that photons can be responsible for the photodesorption of this molecule, in line with other experimental works like Öberg et al. (2009).

Andersson & van Dishoeck (2008) used classical Molecular Dynamics methods with analytical potentials to calculate the photodissociation and photodeorption of water ice. They showed that water is more likely to be dissociated than to photodesorb, hence the discrepancy in the water photodesorption measurements is found in the literature. They distinguished three kinds of mechanism for water photodesorption: kick out of a water molecule by a hot H atom, desorption by the excess of energy during recombination of H and OH, and the desorption of H and OH separately. They reported no evidence of photodesorption below the fifth monolayer. Assuming that water molecules detected in the QMS are photodesorbing, we calculated the photodesorption yield per absorbed photon for the first five monolayers using equation 1 and the average VUV absorption cross-section of solid H₂O and D₂O (see Table 3). We cannot distinguish by monolayer nor by mechanism, hence we show the total water photodesorption.

Andersson & van Dishoeck (2008) affirmed that the more mobility of the photoproducts the more probable they will react with other species. Also, that even they did not include O atoms in their calculations, photoproducts like O₂ could be photodesorbing from the ice. These photoproducts, and the fact that mobility is increased by temperature, could explain the increase of water photodesorption with temperature and its fluence dependence.

Theoretical calculations of Andersson & van Dishoeck (2008) predicted a total photodesorption yield of 5.1 × 10⁻⁴ absorbed photons⁻¹, which is two orders of magnitude lower than our result at 8 K, 7 × 10⁻² molecules per absorbed photon. We are not sure why this difference but QMS data show a higher photodesorption of water related to OH where Andersson & van Dishoeck's (2008) results show the opposite. Speculating, this could be an effect of the UV field spectrum (UV wavelength coverage of the lamp) and/or the fact that photoproducts different than H and OH are photodesorbing and could help in the photodesorption of water by the kick-out process and/or lowering the binding energy of the water molecule about to photodesorb. However, this has to be tested by models.

Molecular dynamics simulations by Arasa et al. (2010, 2011) studied the UV photodissociation of amorphous water and deuterated water ice at 10, 20, 30, and 90 K analysing the effect of ice temperature. They found an increase in water photodesorption of up to 31  per cent for H₂O and 44  per cent for D₂O at 90 K. In our case, we found the similar behaviour but with a higher percentages, 92 and 114  per cent, respectively, at 90 K. Comparing their photodesorption yields with Table 3, a discrepancy of more than two orders of magnitude is observed. As with Andersson & van Dishoeck (2008) photodesorption yields, this could be an effect of the photodesorbing photoproducts. Nevertheless, Arasa et al. (2010, 2011) had short time-scales in their simulations, where thermally activated processes like diffusion and thermal desorption were not probed, arguing that if longer time-scales are considered, a stronger dependence on ice temperature would be expected.

Table 3.

Total photodesorption yield per absorbed photon at different temperatures.

Temp.	H2O	D2O
(K)	(molec. photon−1)
8	0.07	0.05
30	0.08	0.06
50	0.09	0.07
70	0.14	0.12
90	0.15	0.11

Watanabe, Toshikazu & Kouchi (2000) focused on the production of D₂ by the UV irradiation of D₂O ice mantles at 12 K and temperatures above it. They found that D₂ desorption increases drastically for temperatures above 20 K, and this corresponds to the evaporation temperature of D from the ice (Laufer, Kochavi & Bar-Nun 1987). This is in agreement with our results; we also detected this increase between 8 and 30 K (see Fig. 6). Watanabe, Toshikazu & Kouchi (2000), after a dose of 10¹⁸ photons cm⁻², measured a production of D₂ of about 1–2  per cent of the initial number of D₂O molecules at 12 K. In our case, taking the ice thickness for D₂O at 8 K in Table 1 (245 ML) and the total number of photodesorbed D₂ molecules (1.94 ML) for the same experiment, we have a total of ∼ 1 per cent. In agreement with the work of Watanabe, Toshikazu & Kouchi (2000).

The UV filed experienced by the ice mantles in a dark cloud is induced by CRs and is about 10⁴ photons cm⁻² s⁻¹ (see Gredel, Lepp & Dalgarno 1989 and Shen et al. 2004). Using this UV flux in equation 1, accounting for the top five monolayers, the absorbed photons can be calculated. Taking the photodesorption yield for H₂O in Table 3 at 8 K, we calculated a total water desorption rate of ∼12 molecules cm⁻² s⁻¹. For a dark cloud with a mean lifetime of 1 Myr, the total photodesorbed water would be 3.8 × 10¹⁴ molecules cm⁻² ∼0.4 ML. This is a low number compared to gas-phase water abundance towards massive protostars ∼ 10¹⁸ molecules cm⁻² (Boonman et al. 2000; Boonman & van Dishoeck 2003) but in cold clouds, Zmuidzinas et al. (1995) and Tauber et al. (1996) have found that the gas-phase H₂O abundance is low, ∼10⁻⁸–10⁻⁷ related to H₂.

Yeghikyan (2017), using the Cloudy code (Ferland et al. 2013), calculated the effect of the irradiation on the abundance of water (ice and gas) in planetary nebulae (PNs) for 1000 yr. They concluded that this abundance depends on the ionization rate of hydrogen by energetic particles like cosmic rays (CRs). They calculated the VUV flux inside the PN, on average 1.6 × 10⁻³ erg cm⁻² s⁻¹ with an energy per photon close to 10 eV. This gives an average photon flux of 1 × 10⁸ photons cm⁻² s⁻¹. Taking this flux and making the same calculations as before, we found an average water photodesorbed column density of 3.7 × 10¹⁵ molecules cm⁻² in 1000 yr. This result is in line with Yeghikyan (2017) for a CR rate of 10⁻¹³–10⁻¹⁴ s⁻¹ in their models.

Abdulgalil et al. (2017) bombarded water ice with low-energy electrons. They found similar results by detecting the photodesorption of H₂O/D₂O as well as the photoproducts H₂/D₂ and O₂. As in our case, their major photoproduct by far was hydrogen followed by oxygen.

Westley et al. (1995a,b) deposited a 0.5 microns ice, a much thicker ice compared to ours; this could justify the difference in photoproduct formation. They used a hydrogen lamp with a strong Ly-α emission and negligible molecular hydrogen emission around 7.8 eV (besides Ly-α, our lamp presents this bands). They gave a range for the photon flux, from 0.5 × 10¹⁴ to 5 × 10¹⁴ photons/cm² s (ours has been well constrained to 2 × 10¹⁴ photons/cm² s). In addition, their irradiation dose is much higher than ours (∼6 × 10¹⁸ compared with 3.6 × 10¹⁷, about 16 times higher). This could be because they needed a higher dose to remove enough water to be able to measure the thickness more accurately. Nevertheless, we obtained similar results once they are calibrated and properly compared. In their UV irradiation experiments, Westley et al. (1995a,b) observed an incubation stage with incubation dose of ∼10 × 10¹⁸ photons/cm². This suggests a dependence of the photodesorption yield with the UV fluence. Their irradiation experiments are at temperatures above 30 K. Above this temperature H₂ and O₂ can move with a higher freedom than at 8 K as shown in Figs 6 and 8, increasing in the QMS signal with each irradiation period. For these temperatures, water photodesorption gets UV-fluence dependent. Fig. 5 resembles Fig. 2 in Westley et al. (1995b), meaning that their incubation dose corresponds to the higher diffusion of H₂ and O₂ which helps water to photodesorb. Their dependence with UV fluence is in agreement with our results.

Westley et al. (1995a,b) used quartz crystal microbalance to measure the photodesorption yield of H₂O, therefore the yield they derived included all photodesorbing species, namely H₂, OH, H₂O, and O₂. This statement is briefly mentioned by Andersson & van Dishoeck (2008), stating that the removal of H₂O from the surface was driven by H and OH desorption in the top two monolayers. In Fig. 3, if we add yields of H₂, OH, and O₂ together (like in Fig. 9), the trend of total desorbing molecules/radicals will be fluence-dependent which agrees with Westley et al.’s (1995a,b) report, because quartz crystal microbalance can only monitor total loss yield, and it cannot distinguish the proportion of desorbing molecules/radicals. In contrast, our study can provide photodesorption yield of each measurable desorbing species from H₂O ice mantle.

Figure 9.

Photodesorption yield of water and deuterated water found by other groups compared with our work.

Using the yields shown in Table 2, we compared our photodesorption of water with Westley et al. (1995a) by adding the photodesorption yields of OH, H₂O, and O₂ (please confront black asterisks with yellow diamonds in Fig. 9). The photodesorption increases with temperature with a rate similar to Westley et al. (1995a), although H₂ was not added (since it is affected by the isotopic effect explained above in the case of H₂O and D₂O) which means the behaviour could be different. Although we get similar results as Westley, our analytical tools are different. Photolysis becomes increasingly important as temperature increases. This suggests a gas-phase water gradient that increases towards hotter parts in a protoplanetary disc having a direct impact on the position and extension of the water snow line and ultimately affecting the planetary formation and the habitable zone (see Cieza et al. 2016). In addition to photodesorption, CR sputtering is expected to contribute significantly to the release of molecules from ice mantles in cold inter- and circumstellar regions (Dartois et al. 2013).

Öberg et al. (2009) reported the photodesorption yield for D₂O using the RAIR spectra of its stretching band at different temperatures. This technique has the disadvantage that it has no direct relation to column density and it is affected by the OH production. On one hand, they calculated a photodesorption yield ∼3 times higher compared with our D₂O photodesorption yield, with a total 60  per cent error in their calculations. On the other hand, their dependence with temperature is similar to ours, an almost linear increment. The discrepancy between the two yields is probably due to the fact that we measured the D₂O photodesorption yield directly from the QMS, while in Öberg et al. (2009), IR spectroscopy is used, thus leading to an indirect calculation of the D₂O photodesorption yield, since both photodissociation and photodesorption are taken into account in that case. Their H₂O-to-OH ratio is 1 at low temperatures and 2 at high temperatures. We obtained ratios of 2 and 2.5 at 8 and 90 K, respectively. The primary difference is the techniques used by the two works. We consider RAIR spectroscopy a powerful tool but it fails to quantify the column density of the deposited sample, a key parameter to determine a photodesorption yield.

De Simone et al. (2013) address the water photodesorption problem by irradiating water ice at 108 K using a monochromatic light with a wavelength centred at 157 nm. They used time-of-flight mass spectrometer and resonance-enhanced multiphoton ionization to measure the photodesorption cross-section of water at 108 K, on average (6.9 ± 1.8) × 10⁻²⁰ cm² for >10 Langmuir water exposure. They did not provide a photodesorption yield because their relation between water signal and thickness was unknown. Nevertheless, they estimated a yield of 1.8 × 10⁻⁴ molecules per photon. Being this a yield value obtained at high temperature, we can only compare it with our photodesorption yield at 90 K, 2.5 × 10⁻³ molecules per photon for H₂O and 1.5 × 10⁻³ molecules per photon for D₂O. DeSimone et al.’s (2013) estimation is then one order of magnitude lower than ours, Öberg et al. (2009), and Arasa et al. (2011). This is a similar problem as above, knowing that the column density is crucial for measuring the photodesorption yield.

5 CONCLUSION

Photodesorption yields have been given using a novel approach. Based on the experiments presented in this work, we estimated the photodesorption rates at different temperatures for H₂O/D₂O, along with H₂/D₂, OH/OD, and O₂ photoproducts (see Table 2). In the case of O₂, the photodesorption increases with irradiation time, a process highly influenced by the building-up and diffusion of O₂ molecules in the bulk of the ice sample which eventually reach the surface and desorb. This denotes a fluence dependence in the photodesorption of O₂ molecules. This diffusion affects directly other molecules like H₂O, increasing the photodesorption of it. The photodesorption effect can explain the presence of gas-phase water in cold regions where thermal processes are inhibited. Data reported here can be used as inputs for numerical simulations of those regions. We compared our results with existing data and gave possible explanations for the discrepancies. Comparison between our work and previous theoretical simulations is provided for an informational purpose only, since this is beyond the scope of this work.

Acknowledgements

This research was financed by the Spanish MINECO under project AYA2011-29375, AYA2014-60585-P. E. Moreno was also financed by CONSOLIDER grant CSD2009-00038. This work was partially supported by MOST103-2112-M-008-025-MY3 (Y-JC), Taiwan.

APPENDIX A: DATA REDUCTION

In this section, we will explain in detail the steps taken to reduce the raw data from the QMS presented here. We will not explain, however, the calibration of the QMS because as mentioned in the Methods section, it was already explained in a previous paper (see Martín-Doménech et al. 2015).

The main contaminant inside the UHV chamber is hydrogen, followed by water and carbon monoxide (see Fig. A1). Aiming to correct the data from background water contamination, we carried out two irradiations of half an hour before starting the UV photolysis of the ice sample. First, at 8 K, we irradiated the KBr window with no deposition. Then after deposition, we turned the cold finger 90° to irradiate its shield. During deposition, most of the water vapor will condensate on top of the window but a small amount of water molecules will end up sticking elsewhere. These irradiation gave us two different signals for the possible contamination coming from background water (see Fig. A2). On average, water contamination coming from the blank irradiation has an intensity of 3 × 10⁻¹³ A in the QMS while the irradiation of the shield reaches an average intensity of 5 × 10⁻¹³ A, after baseline correction (see below). Also, we used Fig. 1 to remove the contribution of the fragmentation of H₂O and D₂O in the filaments of the QMS. For H₂/H₂O fragmentation ratio, we found a contribution of 1  per cent, 0.7  per cent for D₂/D₂O, 37  per cent for OH/H₂O, and 56  per cent for OD/D₂O. We decided to sum the contributions and subtract the result from the water irradiation experiments. This method was used to correct the m/z signal coming from H₂/D₂ and OH/OD photodesorbed from the ice as well.

Figure A1.

Contamination inside ISAC. Spectrum taken at 8 K and normalized by the most intense mass fragment: m/z = 2.

Figure A2.

Raw data showing the irradiation test to detect the background water contamination in the experiments. Dashed line shows the beginning and the end of the irradiation period for the two test irradiation.

Data in Figs 3, 5, 6, and 8 were corrected by baseline. This means, we brought to zero the QMS signal for clarity, and Fig. A2 is an example of raw data. Nevertheless, quantifications were performed using the raw data after water subtraction, as explained above.

Error estimations were calculated using measurements of the average QMS noise, extracted from the raw data, and the uncertainty in the column density calculation, N, using equation 2. Using the variance formula for the error propagation, we have:

\begin{eqnarray} \delta (Y) = \sqrt{\frac{\delta _i^2 + \delta _j^2 + \delta _k^2 +\cdots + \delta _n^2}{n-1}}. \end{eqnarray}

(A1)

being δ(Y) the error in the photodesorption yield and δ_n the number of possible errors to be considered as the calculation of the column density, UV flux, etc. N, the column density, itself is defined by the error propagation of equation 2 and by the QMS signal noise, which sometimes can be as intense as the measured signal. In the case of the UV flux it depends on the method used to derive it. Muñoz Caro et al. (2010) have a full appendix on how they derived this value. Also, the lamp fluctuations were noticed in the QMS data. The latter was caused by variations on the H₂ pressure inside the UV lamp. Co-adding data would work, for instance, in the acquisition of IR data, but it is not a valid option in this case since the QMS signal and the lamp fluctuations vary from one day to another, it adds uncertainty. H₂O photodesorption measurements in the gas phase are difficult to obtain, hence we improved as much as we could to obtain optimum results.