An uppermost haze layer above 100 km found over Venus by the SOIR instrument onboard Venus Express

Abstract

The Solar Occultation in the InfraRed (SOIR) instrument onboard Venus Express was designed to measure the Venusian atmospheric transmission at high altitudes (65–220 km) in the infrared range (2.2–4.3 μm) with a high spectral resolution. In this work, we investigate the optical properties of Venus’s haze layer above 90 km using SOIR solar occultation observations. Vertical and latitudinal profiles of the extinction coefficient, optical thickness, and mixing ratio of aerosols are retrieved. One of the most remarkable results is that the aerosol mixing ratio tends to increase with altitude above 90 km at both high and low latitudes. We speculate how aerosols could be produced at such high altitudes.

Introduction

The clouds above Venus consist of a main cloud deck located between approx. 47 and 70 km surrounded by thinner hazes above and below. The upper haze layer was observed at altitudes as high as 90 km (Esposito et al. 1983). The presence of high-altitude hazes/aerosols was also inferred from the analysis of limb profiles observed by the Venus Monitoring Camera (VMC) onboard ESA’s Venus Express (Vex) (Limaye et al. 2015) and from SPICAV-IR and UV solar occultations (Luginin et al. 2016). The mean particle radius (~ 1 μm) within the cloud layer was previously determined from polarimetry observations in the early 1970s (Hansen and Hovenier 1974). These particles are most likely composed of concentrated sulfuric acid (Esposito 1984), as deduced from refractive index measurements. The existence of smaller particles within and above the clouds with an effective radius of 0.23 ± 0.04 μm and a refractive index of 1.45 ± 0.04 was suggested from Pioneer Venus Orbiter (PVO) Cloud Photopolarimeter measurements at 550 nm (Kawabata et al. 1980). From PVO Cloud Photopolarimeter observations, Sato et al. (1996) reported that these haze particles have an effective radius of 0.25±0.05 μm and a refractive index of $1.435 \pm 0.02$ in the North Pole region; in the South Pole region, these haze particles have an effective radius of 0.29 ± 0.02 μm and a refractive index of $1.45 \pm 0.02$. The haze optical thickness above the clouds in the polar regions was found to be equal to 0.8 at 365 nm but only 0.06 at low latitudes (Kawabata et al. 1980). In Takagi and Iwagami (2011), the cloud optical thickness, which includes the optical thickness of the haze layer, was estimated from data collected by the previous entry probes, such as the Soviet Venera series; the haze optical thickness from 70 to 90 km at 920 nm was found to be 1.0. Vertical extinction profiles were obtained from data acquired over several wavelengths in the UV and IR domains with SPICAV/SOIR instruments (Wilquet et al. 2009). The extinction values differ according to different wavelengths. The optical properties of aerosols have been inferred from their wavelength dependence. Long-term temporal variations in the haze layer were investigated using PVO cloud photopolarimeter observation data (Braak et al. 2002); a gradual decrease in the haze particle column density was observed during the Pioneer Venus mission. The haze optical thickness was found to be 0.25 in 1978 and 0.1 or less in 1990. From PVO Cloud Photopolarimeter observation data, Sato et al. (1996) also showed short-term temporal variations in the haze optical thickness in the polar regions. As shown in Belyaev et al. (2008), after a sharp increase in the SO₂ abundance between 1967 and 1979, a gradual decrease in the SO₂ abundance beginning in 1995 was observed. Several possible explanations of such variations were later considered by Mills and Allen (2007). More recently, another global increase in the SO₂ abundance between 2004 and 2007 was observed (Belyaev et al. 2008). Esposito et al. (1988) reported a correlation coefficient of 0.8 between the SO₂ abundance and polar haze optical thickness from PVO Cloud Photopolarimeter observations. However, it is still unclear how hazes are produced and what their composition is.

The haze optical properties up to 90 km were presented in Wilquet et al. (2009, 2012), who also described an analysis of spectra from the SOIR instrument, a high-resolution infrared echelle spectrometer onboard Venus Express. Wilquet et al. (2012) reported that the aerosol extinction coefficient is significantly smaller at high latitudes (at least by a factor 10) than in the equatorial region. Wilquet et al. (2009) also showed the existence of at least two types of particles with radii of ~ 0.1–0.3 μm and ~ 0.4–1.0 μm depending on the altitude. On the other hand, the SO and SO₂ mixing ratios were shown to increase with the altitude from 85 to 105 km (Belyaev et al. 2012; Mahieux et al. 2015b). These observations were tentatively explained by the existence of a still unknown source of SO and SO₂ at high altitudes. One possible source could be the photodissociation of SO₃, which results from the evaporation of H₂SO₄ droplets. For example, Zhang et al. (2012) showed important chemical pathways for sulfur species related to aerosols. Additionally, it has been speculated that aerosols and sulfur compounds are connected by condensation and evaporation (Zhang et al. 2012). However, upper limit measurements of H₂SO₄ using submillimeter ground-based observations make this suggestion unlikely (Sandor et al. 2012). Clearly, what occurs above 90 km is not yet understood.

Venus Express, the first European mission to Venus, was intended to provide a global investigation of the atmosphere and plasma environment above Venus (Svedhem et al. 2007). The spacecraft was launched on November 2005 from Baikonur, Kazakhstan, arrived at Venus on April 2006, and continued operating for more than 8 years. It was inserted in an elliptical orbit; its perigee was above the North Pole ($\sim 250\,\,{\text{km}}$), and its perigee was above the South Pole ($\sim 66,000\, {\text{km}}$). The satellite’s orbital period was 24 Earth hours. Several instruments on the Venus Express spacecraft were able to observe the Venus atmosphere over a broad range of wavelengths. VMC images have also been used to infer some results on hazes via estimates of the slant unit optical depth from limb determinations at four visible/near-infrared wavelengths (Limaye et al. 2015). The Visible and Infrared Thermal Imaging Spectrometer (VIRTIS-M IR) (de Kok et al. 2011) and Spectroscopy for the Investigation of the Characteristics of the Atmosphere of Venus/Solar Occultation at InfraRed (SPICAV/SOIR) (Wilquet et al. 2009) instruments were able to investigate the upper haze layer above the clouds on the night side. The aim of this study is to examine the haze optical properties above that altitude by analyzing SOIR data to understand the processes at play at such high altitudes.

Description of the observations

The SOIR instrument

SOIR, one of the three channels of the SPICAV/SOIR instrument (Bertaux et al. 2007) onboard Venus Express, is a compact and high-resolution IR echelle grating spectrometer working at high diffraction orders (simply called orders hereafter) between 101 and 194 (Nevejans et al. 2006). More details on the instrument can be found in Nevejans et al. (2006), Mahieux et al. (2008, 2009) and Vandaele et al. (2013); only a summary is given here. The SOIR instrument operates in the IR domain [2.2–4.3 μm (2200 cm⁻¹–4400 cm⁻¹)], and its spectral resolution varies between $0.1$ and $0.19\,{\text{cm}}^{ - 1}$. The free spectral range is equal to $24 \,{\text{cm}}^{ - 1}$. An acousto-optical tunable filter (AOTF) was placed beyond the optical entrance of the instrument to select the echelle diffraction orders to be measured during an occultation, and it was designed to have a full width at half maximum (FWHM) of $24\, {\text{cm}}^{ - 1}$. Four orders are recorded simultaneously during an occultation. The detector pixels are summed on board in the spatial direction to form two bins; this means that the spectral resolution is lowered by a factor of two by binning. The vertical resolution varies between 200 and 700 m when measuring at high northern latitudes and between 2 and 5 km in the Southern Hemisphere (Mahieux et al. 2010). The observed spectra at northern latitudes have a high vertical resolution compared to those observed at southern latitudes due to the elliptical orbit of the satellite with its perigee above the North Pole.

SOIR always operated in solar occultation mode, i.e., the instrument’s line of sight (LOS) was pointed towards the Sun. Figure 1 shows the geometry of a solar occultation observation. Solar occultation is a technique in which the spectra are obtained by dividing the spectra measured while looking at the Sun through the Venus atmosphere by an exoatmospheric reference spectrum (Vandaele et al. 2008, 2013) (ex. Fig. 2). The tangent altitude of each measurement is defined as the minimum distance between the surface of Venus and the instrument line of sight. As Vex/SOIR moved along its orbit, successive tangent altitudes were defined. SOIR acquired observations either at sunset or at sunrise on a 1-s time cycle. The instrument probed the Venus atmosphere at the terminator, which is a unique region, since it represents the boundary between the dark and illuminated sides of the planet. The local solar time of each observation was always 6 AM or 6 PM, because the evening occultation occurred when the Sun was setting (as seen from SOIR) and the morning occultation occurred when the Sun was rising.

Fig. 1

Geometry of solar occultation measurements and the analysis method in this study (modified from Fig. 4 of Vandaele et al. 2008). In the ingress case depicted here, the instrument is taking a measurement every second. The tangent altitude $h_{tg, j}$ defines the layers in the atmosphere. In each layer $L_{i}$, the temperature $T_{i}$ and pressure $P_{i}$ are assumed to be constant

Fig. 2

Example of observed transmittances (Dec. 03, 2007, order 147, orbit 591, latitude $9^\circ {\text{S}}$, longitude $209^\circ$) above an altitude of 95 km. These wavelengths are selected to visualize the observed spectra. The observed transmittances above 122 km are almost unity, because the solar light is not yet being absorbed by the atmosphere. At altitudes lower than 122 km, the standard deviation on the observed transmittance is small enough to isolate an individual observed transmittance

SOIR transmittance

At the beginning of an occultation, the solar light path does not traverse the Venus atmosphere. SOIR starts recording solar spectra at the outer region of the atmosphere to obtain at least 40 spectra and to define the reference Sun spectra. For tangent altitudes lower than 220 km, the observed transmittances are calculated by dividing the spectrum recorded at the current time by the reference Sun spectrum. At lower altitudes, the instrument line of sight penetrates deeper into the atmosphere, and continuum absorption—over an order width—occurs due to haze and atmospheric molecules such as CO₂. At the end of an occultation, the signal becomes zero, because the solar light is completely absorbed by the clouds and atmospheric species.

Figure 2 gives an example of the observed transmittances during one occultation (orbit 591.1, Dec. 3, 2007, order 147, latitude $9^\circ {\text{S}}$, longitude $209^\circ$). As shown in Fig. 2, the observed transmittances above 122 km are almost unity within the noise, because the solar light is not yet absorbed by the atmosphere. On the other hand, the difference between the spectra recorded at 121.02 km and 115.54 km is 0.33%, while the standard deviation at such altitudes is on the order of 0.04%. The standard deviation of an observed transmittance is calculated as:

$$ S = \sqrt {\frac{1}{n}\mathop \sum \limits_{i = 1}^{n} \left( {x_{i} - \bar{x}} \right)^{2} } , $$

(1)

where $n$ is the number of spectra at a particular altitude, $x_{i}$ represents the observed values, and $\bar{x}$ is the mean value of the observations at that altitude. Additionally, systematic errors, such as those due to pointing errors, may be canceled out when the spectral ratio is calculated, because an occultation is basically a relative measurement. The methods used to estimate the pointing drift and detector sensitivity drift are described in “Appendix”.

It is found that the standard deviation of the observed transmittance is small enough to isolate an individual observed transmittance regardless of the high altitudes (above 90 km) of the observations. In this work, we will work with spectra obtained at altitudes between 70 and 120 km; the effect of molecular spectral absorption has been removed from these spectra (see “Removal of the molecular absorption effect” section for details on this procedure).

Figure 3 and Table 1 show the latitudinal distribution of all the observations considered in this work. These observations are also listed in Table 2 (morning occultations) and Table 3 (evening occultations). The orders that have no strong CO₂ absorption bands are considered, whereas the orders covering the CO₂ band at 3 μm (orders 152 to 169) and 3.7 μm (orders 101 to 110) are not considered.

Fig. 3

The latitudes of selected SOIR occultations are plotted as a function of the VEX orbit number. The corresponding first days of each Earth year are given at the top of the figure. The dots (●) are for morning solar occultations, while the crosses (×) are for evening solar occultations. There is one orbit per Earth day. There are 128 occultations considered in this work. The data cover the period from 2006 to 2010

Table 1 Latitudinal distribution of the occultations assuming hemispheric symmetry

Table 2 Morning occultations employed in this study

Table 3 Evening occultations employed in this study

Method of analysis

Removal of the molecular absorption effect

The gas transmittance $T_{\text{gas}}$ due to atmospheric molecules such as CO₂, H₂O, HCl, and HF, which have absorption lines in the orders considered in this work, is removed from the observed transmittance $T_{\text{obs}}$. $T_{\text{gas}}$ is calculated as:

$$ T_{\text{gas}} = e^{{ - \tau_{\text{gas}} }} , $$

(2)

where $\tau_{\text{gas}}$ is the total optical thickness due to all species obtained as:

$$ \tau_{\text{gas}} = \mathop \sum \limits_{i} \tau_{i} , $$

(3)

where $\tau_{i}$ is the optical thickness of species $i$ ($i$= CO₂, H₂O, HCl, and HF), integrated along the full line of sight (LOS):

$$ \tau_{i} = \sigma_{i} \int\limits_{\text{LOS}} {n_{i} \,{\text{d}}s} , $$

(4)

where $\sigma_{i}$ and $n_{i}$ are the absorption cross section and the number density of species $i$, respectively. The absorption cross sections are calculated using a line-by-line method taking the CO₂ parameters from the HITEMP database (Wattson and Rothman 1992) and the other parameters from the HITRAN 2012 databases (Rothman et al. 2013). The initial number densities and temperature and pressure profiles are taken from the Venus International Reference Atmosphere (VIRA) model (Seiff et al. 1985). Only the absorption lines present in the central order are considered in the calculation of the synthetic spectra. As an approximation, we do not consider the diffraction order addition procedure, because the AOTF bandpass function is wider than the spectral width of the detector, implying that more than one diffraction order is measured by the detector (Mahieux et al. 2012). This approximation is valid here, since the value of the AOTF transfer function is negligible in the adjacent orders when compared to the central order since we considered only the main orders, for which the absorption structures in the adjacent orders are weak [see Mahieux et al. 2015a (the paper about the rotational temperature)]. $T_{\text{gas}}$ is fitted by adjusting the number density of the gas species and convolving the absorption cross section by a triangular filter ($0.14 {\text{cm}}^{ - 1}$ FWHM). An example of $T_{\text{gas}}$ and $T_{\text{obs}}$ is given in Fig. 4 for a typical observation (Sep. 05, 2006, order 149, orbit 137, latitude $79\,{\text{N}}^\circ$, longitude $237^\circ$, altitude 110.3 km). $T_{\text{haze}}$ is obtained by dividing $T_{\text{obs}}$ by $T_{\text{gas}}$ at each observed altitude as:

$$ T_{\text{haze}} = \frac{{T_{\text{obs}} }}{{T_{\text{gas}} }} $$

(5)

Fig. 4

Example of a calculated transmittance ($T_{\text{gas}}$, black spectrum), an observed transmittance ($T_{\text{obs}}$, red spectrum), and a residual transmittance ($T_{\text{haze}}$, dotted blue spectrum) measured at an altitude of 110.3 km corresponding to the observation obtained on Sep. 05, 2006 (order 149, orbit 137.1, latitude $79\,{\text{N}}^\circ$, longitude $237^\circ$). These wavelengths are selected to visualize the analysis method

Retrieval of the haze optical properties

Each solar occultation measurement can be assigned to a specific altitude at the limb. As shown in Fig. 1, the tangent altitude $h_{tg}$ decreases with time during the occultation in the case of an ingress. The tangent altitude $h_{tg, j}$ corresponding to the jth measurement defines an atmospheric shell. The region between two successive measurement altitudes $h_{tg, j - 1}$ and $h_{tg, j}$ defines the jth layer ($L_{j}$).

$\tau_{j}$, the horizontal optical thickness corresponding to layer $L_{j}$ and all the above-located layers, is defined as:

$$ \tau_{j} = - \ln \left( {T_{{{\text{haze}}, j}} } \right), $$

(6)

where $T_{{{\text{haze}},j}}$ is defined in Eq. (5) for the jth observation path.

$\tau_{j,in}$, the horizontal optical thickness of layer $L_{j}$, is obtained considering the so-called onion peeling method and can be written as:

$$ \tau_{j,in} = \tau_{j} - \mathop \sum \limits_{i = 1}^{j - 1} 2\,{\text{d}}x_{i} \times k_{i} , $$

(7)

where $k_{i}$ is the aerosol extinction coefficient of layer $L_{i}$ and $dx_{i}$ is the horizontal path length in layer $L_{i}$, which is outside relative to layer $L_{j}$.

We define the local extinction coefficient $k_{j}$, calculated by dividing $\tau_{j,in}$ by the local length $2\,{\text{d}}x_{j}$, as:

$$ k_{j} = \frac{{\tau_{j, in} }}{{2\, {\text{d}}x_{j} }}, $$

(8)

where the factor 2 comes from the symmetry on both sides of the tangent point.

The vertical distribution of extinction coefficients is obtained by repeating the process described above. Figure 5 (top) is an example of the vertical distribution of the extinction coefficient. The extinction error $\Delta k$ is estimated as:

$$ \Delta \tau_{j} = \frac{{\Delta T_{{{\text{haze}}, j}} }}{{T_{{{\text{haze}},j}} }} $$

(9)

$$ \Delta \tau_{j, in} = \sqrt {\left( {\Delta \tau_{j} } \right)^{2} + \left( {\mathop \sum \limits_{i = 1}^{j - 1} \left( {2{\text{d}}x_{i} \times \Delta k_{i} } \right)} \right)^{2} } $$

(10)

$$ \Delta k_{j} = \frac{{\Delta \tau_{j,in} }}{{2{\text{d}}x_{j} }}, $$

(11)

where $\Delta T_{{{\text{haze}}, j}}$ is the standard deviation of $T_{{{\text{haze}},j}}$, calculated using Eq. (1), and $\Delta T_{{{\text{haze}}, j}}$ is estimated to be 0.1%, whereas the average $\Delta k_{j}$ for all observations is estimated to be almost 30% at all observation paths. $\Delta k_{j}$ becomes much larger than $\Delta T_{{{\text{haze}},j}}$ because of the errors accumulated layer by layer through the sum in Eq. (10).

Fig. 5

Example of a vertical profile of extinction coefficients (top panel) and the corresponding normalized extinction coefficients (bottom panel) for orbit 137, order 149 (Sep. 05, 2006, latitude $79\,{\text{N}}^\circ$, longitude $237^\circ$). The average errors are almost 26% for the extinction coefficient and 42% for the normalized extinction coefficient

The vertical optical thickness $\tau_{\text{vert}}$ is calculated by integrating the extinction vertically as:

$$ \tau_{\text{vert}} = \smallint k_{j} {\text{d}}z . $$

(12)

The normalized extinction $m_{j}$ (in dimensions of the mixing ratio) is defined as:

$$ m_{j} = \frac{{k_{j} }}{{n_{{{\text{CO}}_{2} }} \cdot S}}, $$

(13)

where $n_{{{\text{co}}_{2} }}$ is the total CO₂ number density in layer $L_{j}$ and $S$ is the extinction coefficient cross section. The normalized extinction coefficient is a nondimensional quantity. The total CO₂ number density is obtained from previous studies (Mahieux et al. 2010, 2012, 2015a). $S$ is the CO₂ Rayleigh scattering cross section at 3.0 μm, which is expected to be equal to $1.1 \times 10^{ - 29} \left[ {{\text{cm}}^{2} } \right]$, extrapolated from the values in the visible region (in Table 8-2 of Tohmatsu and Ogawa 1990). Figure 5 (bottom) is an example of the vertical distribution of the normalized extinction coefficient. The errors of $\tau_{\text{vert}}$ and $m_{j}$ are estimated as:

$$ \Delta \tau_{\text{vert}} = \mathop \sum \limits_{i = 1}^{j} (\Delta k_{i} \times {\text{d}}z) $$

(14)

$$ \Delta m_{j} = \sqrt {\left( {\frac{{\Delta k_{j} }}{{n_{{{\text{co}}_{2} }} \cdot S}}} \right)^{2} + \left( {\frac{{k_{j} \cdot \Delta n_{{{\text{co}}_{2} }} }}{{(n_{{{\text{co}}_{2} }} )^{2} \cdot S}}} \right)^{2} } , $$

(15)

where $\Delta n_{{{\text{CO}}_{2} }}$ is the error on the CO₂ number density, which was obtained from previous studies (Mahieux et al. 2010, 2012, 2015a). The average $\Delta m$ for all observations is estimated to be 40% for all observations.

Results

Extinction profiles

The average extinction coefficient profiles are plotted in Fig. 6 (for the morning and evening sides of the terminator and for high and low latitudes). As shown in these profiles, we observe that hazes also appear to be present at altitudes above 90 km. Moreover, the extinction coefficient profiles show an obvious change in the slope at $\sim 95 {\text{km}}$. Finally, the extinction coefficients at low latitudes are larger than those at high latitudes, and this ratio is smaller than that already reported by Wilquet et al. (2012). In the 70–90 km altitude region, the average vertical optical thickness is equal to 0.19 and 0.24 for the high- and low-latitude regions, respectively. In the 90–110 km altitude region, the average vertical optical thickness is equal to 0.005 and 0.015 for the high- and low-latitude regions, respectively. The mean variability for all the observations is 12% for the 70–90 km region and 14% for the 90–110 km region. The optical thickness at low latitudes is larger than that at high latitudes. However, Kawabata et al. (1980) showed that the haze optical thickness is larger at high latitudes than at low latitudes. Braak et al. (2002) reported a long-term temporal variation in the haze optical thickness, which decreased by a factor of 0.3 between 1980 and 1992. The high-latitude optical thickness was relatively large in the recent past.

Fig. 6

Average extinction coefficient profiles for morning high-latitude (latitude $> 60^\circ$, black line), evening high-latitude (latitude $> 60^\circ$, red line), morning low-latitude (latitude $< 60^\circ$, blue line), and evening low-latitude (latitude $< 60^\circ$, green line) observations. The average of the variability is 21.5% for the morning high-latitude observations

Normalized extinction coefficients

The average normalized extinction coefficient profiles are plotted in Fig. 7 (for the morning and evening sides and for high and low latitudes). A significant increase in the normalized extinction coefficient with the altitude is observed above 90 km for both the high- and the low-latitude regions. The normalized extinction coefficients for both the morning and the evening occultations at low latitudes are almost one order of magnitude larger than those at high latitudes at altitudes above 90 km.

Fig. 7

Average of all normalized extinction coefficients for morning high-latitude (latitude $> 60^\circ$, black line), evening high-latitude (latitude $> 60^\circ$, red line), morning low-latitude (latitude $< 60^\circ$, blue line), and evening low-latitude (latitude $< 60^\circ$, green line) observations. The average of the variability is 34.3% for the morning high-latitude observations

Discussion

In Esposito et al. (1983), the upper haze layer was observed at altitudes as high as 90 km. Additionally, the presence of high-altitude haze was inferred from VMC images and from SPICAV-IR and UV occultations onboard Venus Express (Limaye et al. 2015; Luginin et al. 2016). We derived the haze optical properties at altitudes above 90 km with SOIR data. We showed that haze is present at such high altitudes and that the normalized extinction coefficient increases above 90 km at both high and low latitudes. The relationship between sulfur compounds and haze has already been discussed by several previous studies (Mills et al. 2007; Zhang et al. 2012). Recently, observations of SO and SO₂ at high altitudes have been reported; for example, SOIR detected an increase in the SO₂ VMR above 85 km at the terminator (Mahieux et al. 2015b). Belyaev et al. (2012) reported the presence of two layers of SO and SO₂: a lower layer between 65 and 80 km observed by SOIR and an upper layer between 85 and 105 km observed by the UV channel of the SPICAV/SOIR instrument. The mixing ratios of SO and SO₂ decrease from 0.2 to 0.02 ppmv between 65 and 80 km and increase from 0.05 to 2 ppmv between 85 and 105 km. We postulate that the increases in the SO and SO₂ mixing ratios between 85 and 105 km and the increase in the haze extinction coefficient between these altitudes may be either coincidental or connected. Such a connection does not seem likely between 65 and 80 km. However, it is still unclear how haze is produced and of what its composition consists. We speculate that sources of SO and SO₂, such as H₂SO₄ or Sx, are transported upwards and are then photodissociated.

The following process is proposed to explain the observed increase in the haze normalized extinction coefficient at high altitudes (above 90 km):

• (1) Sources of haze particles (such as H₂SO₄ or S_X) are transported upward at a velocity larger than the sedimentation velocity from the cloud deck.

• (2) These transported aerosols either evaporate (in the case of H₂SO₄, see Zhang et al. (2012), for example) or react with other compounds (sulfur and oxygen atoms), and SO and SO₂ will be produced at high altitudes. For example, Mills and Allen (2007) and Zhang et al. (2012) proposed detailed formation schemes starting either from H₂SO₄ or S_X to form SO and SO₂.

• (3) Haze can be produced by chemical processes at such high altitudes through a series of chemical reactions initiated in particular by the photodissociation of SO and SO₂, as discussed in Mills and Allen (2007) and Zhang et al. (2012).

• (4) The size of the produced haze particles should be smaller than those of particles transported from their sources for the following reason.

The extinction coefficient is defined as:

$$ k = \sigma \cdot N, $$

(16)

where $\sigma$ is the extinction cross section and $N$ is the number density of the scattering material. When the haze particles become smaller, keeping the haze mass constant, the extinction cross sections decrease, and the number density becomes larger. For example, when the particle size is divided by two, the extinction cross section becomes four times smaller; on the other hand, the number density becomes eight times larger. Under these circumstances, the extinction coefficient becomes larger when the particles are produced by chemical reactions than when the sources of haze are transported from below.

The normalized extinction coefficients can increase at high altitudes because of the processes proposed above. Figure 7 shows that the normalized extinction coefficients at low latitudes are almost one order of magnitude larger than those at high latitudes. This can be explained by the fact that haze is produced in a larger amount at low latitudes than at high latitudes. From the previous discussion, larger aerosols are likely transported to these higher altitudes, while smaller particles are produced there. This explanation could be verified by numerical models. This hypothesis is compatible with the measurements of Wilquet et al. (2009), who demonstrated the existence of at least two types of particles at high altitudes: mode 1 of mean radius $0.1{ \leqq }r{ \leqq }0.3$ μm and mode 2 of $0.4{ \leqq }r{ \leqq }1.0$ μm.

Conclusions

The optical properties of the upper haze layer at altitudes above 90 km were studied in this work. A significant increase in the normalized extinction coefficient was also observed above 90 km at both high and low latitudes, which could be linked to the vertical profiles of SO and SO₂. The processes by which the normalized extinction coefficient increases were discussed, and we proposed the following mechanisms to explain our observations. Aerosols, including sulfur-based compounds, are transported upwards at a velocity larger than the sedimentation velocity from the cloud deck. The transported aerosols then evaporate or react to produce SO and SO₂ at such high altitudes. At high altitudes, haze particles are produced by chemical processes involving SO and SO₂. Since the normalized extinction coefficient increases at high altitudes, we propose that the size of the haze particles that are produced is smaller than those of transported aerosol particles. Haze seems to be produced in a larger extent at low latitudes than at high latitudes, as deduced from the normalized extinction coefficients that are several times larger at low latitudes than at high latitudes.

Appendix

Figure 8 (Vandaele et al. 2013) and Fig. 9 (Trompet et al. 2016) show examples of the signal obtained during one solar occultation. These examples correspond to an egress (Fig. 8) when the tangent altitude increases with time (sunrise), and an ingress (Fig. 9) when the tangent altitude decreases with time (sunset). The top panels of Figs. 8 and 9 show the signal in analog digital units (ADU) measured on one detector pixel during a solar occultation. In the case of Fig. 8, the signal can be separated into three parts: at the beginning, no light reaches the SOIR when Venus Express is in the umbra region (black line); then, the occultation occurs when Venus Express is in the penumbra region (red line); finally, the instrument looks directly at the Sun and measures exoatmospheric raw spectra without atmospheric absorption (blue line). The middle panel of Fig. 8 and the bottom panel of Fig. 9 show the process of determining the signal level in the exoatmosphere. There are significant changes at 30–50 s (Fig. 8) and at 260–280 s (Fig. 9). In these cases, the whole exoatmospheric level is determined with the regression lines for tangent altitudes of 220 km or above (after 130 s in Fig. 8, before 180 s in Fig. 9).

Fig. 8

Panel 1: Typical evolution through time of the signal (in ADU) on one pixel of the detector during a solar occultation. This example corresponds to an egress: at the beginning, no light reaches the detector when VEX is in the umbra region (black line); then, the occultation occurs when VEX is in the penumbra region (red line); finally, the instrument looks directly at the Sun and measures exoatmospheric raw spectra (blue line). The exoatmosphere and umbra zones used to derive the noise are indicated by the boxes. Panels 2 and 3: Variation of the noise signal around the mean value of the two zones (exoatmosphere in Panel 2 and umbra region in Panel 3). The standard deviations are shown by the lines in green (taken from A.C. Vandaele et al. 2013)

Fig. 9

(A) Separation of the Sun region into different zones for one detector pixel as a function of the index, i.e., the time after the beginning of the measurement. Lines corresponding to specific tangent altitudes are plotted. (B) Plot of the fit over the S region and extrapolation in the T region. The graph inside shows the residuals of the fit (taken from Trompet et al. 2016)

Fluctuations in the level due to pointing drift are shown in the middle panel of Fig. 8 and in the enlarged figure in the bottom panel of Fig. 9; in these cases, the raw spectrum are ± 0.1 ADU/108 ADU = ± 0.09% and ± 0.05 ADU/20 ADU = ± 0.25%, respectively. These values are on the same level or less than the difference between the spectra recorded at 121.02 km and 115.54 km (0.33%, Fig. 2 in this work). That is, observed transmittances up to 120 km are significant even if pointing drift is considered.

Fluctuations in the level due to detector sensitivity drift are not found in Fig. 8 (less than 0.1% per 100 s), but are found in Fig. 9 as an inclination of the regression line. The inclination is estimated to be approximately 1% per 100 s in the exoatmosphere and 0.2% per 20 s mainly during observations (260–280 s). These values are on the same level or less than the difference between the spectra recorded at 121.02 km and 115.54 km (0.33%, Fig. 2 in this work). That is, observed transmittances up to 120 km are significant even if detector sensitivity drift is considered. However, other potential sources (e.g., pointing drift) may be included in this study. The above explanation does not evaluate the measurement errors due to pointing drift and detector sensitivity drift directly. However, this estimation of measurement errors should be effective.