+ All Categories
Home > Documents > Quartz-glass chopping method of measuring solar irradiance

Quartz-glass chopping method of measuring solar irradiance

Date post: 01-Oct-2016
Category:
Upload: georg
View: 212 times
Download: 0 times
Share this document with a friend
6
Quartz-glass chopping method of measuring solar irradiance Paul Schlyter and Georg Witt A solar irradiance photometer was launched on a pair of ESRO Centaure rockets. The instrument, a solar irradiance photometer used six wavelength channels. Three of these channels were used to measure meso- spheric ozone by the occultation method. Two other channels were equipped with 214-nm Fabry-Perot fil- ters to monitor the absorption of sunlight required for the excitation of NO y-band fluorescence,which was measured by another instrument in the same payload. One of the 214-nm channels used a quartz lens, while the other used a glass lens. With a glass-quartz chopping technique, stray light with wavelengths longer than 300 nm could be eliminated. Since ozone is absorbing at 214 nm, these two channels were also used for measuring mesopheric ozone. By comparing the ozone measurements from the two 214-nm channels with the measurements from the three channels intended for ozone measurements, one can obtain the perfor- mance of the glass-quartz chopping method. 1. Introduction By using Fabry-Perot interference filters, one can make fairly compact photometers suitable for small meteorological rockets. These filters have two major disadvantages, however: (1) The wavelength response is dependent on the angle of incidence. This can be overcome by placing a diffusor in front of the filter, making the light paths through the filter independent of the direction of the incident light. (2) The filter transmits too much light at the long wavelength side of the peak. This becomes an in- creasingly severe problem the shorter the wavelength for two reasons: (a) It is more difficult to make inter- ference filters with good out-of-band rejection at shorter wavelengths. (b) With the sun as the light source, the intensity decreases rapidly with decreasing wavelength in the UV, which makes it especially difficult to elimi- nate unwanted radiation. The Fabry-Perot filters can be used down to -250 nm without severe problems, since for these wavelengths blocking filters and selective mirrors are available. At still shorter wavelengths, however, such filters will re- quire alternative methods for background rejection. The authors are with University of Stockholm, Meterology De- partment, Arrhenius Laboratory, Fack, S-104 05 Stockholm, Sweden. Received 19 May 1980. 0003-6935/81/142528-06$00.50/0. C 1981 Optical Society of America. II. Instrument The solar irradiance photometer (SIP) is seen in Fig. 1. It is equipped with six different wavelength chan- nels. Each channel has a Suprasil I diffusor, which is irradiated by the sun. Behind each diffusor there is a focusinglens, an interference filter, and a mirror which reflects the light beam onto a photomultiplier tube (PMT). All six light beams are reflected onto the same PMT. The current from the PMT is then amplified in a logarithmic amplifier and then transmitted to the ground. Most of the channels also have a NiSO4 -6H 2 0 crystal in the light path, and three channels also use selective mirrors, while the other three use standard Al mirrors; see Table I, which also shows the manufac- turers of the optical elements. The purpose of the NiSO 4 -6H 2 0 crystal is to filter out visible light. Therefore, such a crystal is not used in the 450-nm visible light channel. Selective mirrors were used in three channels to fur- ther enhance selectivity. The selectivity of the 450-nm channel was so good, however, that a regular mirror could be used instead. And for the two 214-nm chan- nels, selective mirrors for such short wavelengths were not available. To improve the selectivity of the 214-nm interference filter, we used the glass-quartz chopping method and equipped one of the 214-nm channels with a glass lens and the other with a lens of fused silica. The 214-nm channel with the glass lens is 214G and the one with the Suprasil lens 214Q. The diffusors are positioned in such a way that as the rocket is spinning around its axis, the diffusors become exposed to sunlight one by one. The rocket also has a precession, which of course influences the irradiance upon the diffusors (Fig. 2). To monitor this influence, the 450-nm channel is used, since in this wavelength 2528 APPLIED OPTICS / Vol. 20, No. 14 / 15 July 1981
Transcript

Quartz-glass chopping method of measuring solar irradiance

Paul Schlyter and Georg Witt

A solar irradiance photometer was launched on a pair of ESRO Centaure rockets. The instrument, a solar

irradiance photometer used six wavelength channels. Three of these channels were used to measure meso-

spheric ozone by the occultation method. Two other channels were equipped with 214-nm Fabry-Perot fil-

ters to monitor the absorption of sunlight required for the excitation of NO y-band fluorescence, which wasmeasured by another instrument in the same payload. One of the 214-nm channels used a quartz lens, while

the other used a glass lens. With a glass-quartz chopping technique, stray light with wavelengths longer

than 300 nm could be eliminated. Since ozone is absorbing at 214 nm, these two channels were also used for

measuring mesopheric ozone. By comparing the ozone measurements from the two 214-nm channels with

the measurements from the three channels intended for ozone measurements, one can obtain the perfor-mance of the glass-quartz chopping method.

1. Introduction

By using Fabry-Perot interference filters, one canmake fairly compact photometers suitable for smallmeteorological rockets. These filters have two majordisadvantages, however:

(1) The wavelength response is dependent on theangle of incidence. This can be overcome by placing adiffusor in front of the filter, making the light pathsthrough the filter independent of the direction of theincident light.

(2) The filter transmits too much light at the longwavelength side of the peak. This becomes an in-creasingly severe problem the shorter the wavelengthfor two reasons: (a) It is more difficult to make inter-ference filters with good out-of-band rejection at shorterwavelengths. (b) With the sun as the light source, theintensity decreases rapidly with decreasing wavelengthin the UV, which makes it especially difficult to elimi-nate unwanted radiation.

The Fabry-Perot filters can be used down to -250 nmwithout severe problems, since for these wavelengthsblocking filters and selective mirrors are available. Atstill shorter wavelengths, however, such filters will re-quire alternative methods for background rejection.

The authors are with University of Stockholm, Meterology De-partment, Arrhenius Laboratory, Fack, S-104 05 Stockholm,Sweden.

Received 19 May 1980.

0003-6935/81/142528-06$00.50/0.C 1981 Optical Society of America.

II. Instrument

The solar irradiance photometer (SIP) is seen in Fig.1. It is equipped with six different wavelength chan-nels. Each channel has a Suprasil I diffusor, which isirradiated by the sun. Behind each diffusor there is afocusing lens, an interference filter, and a mirror whichreflects the light beam onto a photomultiplier tube(PMT). All six light beams are reflected onto the samePMT. The current from the PMT is then amplified ina logarithmic amplifier and then transmitted to theground. Most of the channels also have a NiSO4 -6H2 0crystal in the light path, and three channels also useselective mirrors, while the other three use standard Almirrors; see Table I, which also shows the manufac-turers of the optical elements.

The purpose of the NiSO4 -6H20 crystal is to filterout visible light. Therefore, such a crystal is not usedin the 450-nm visible light channel.

Selective mirrors were used in three channels to fur-ther enhance selectivity. The selectivity of the 450-nmchannel was so good, however, that a regular mirrorcould be used instead. And for the two 214-nm chan-nels, selective mirrors for such short wavelengths werenot available. To improve the selectivity of the 214-nminterference filter, we used the glass-quartz choppingmethod and equipped one of the 214-nm channels witha glass lens and the other with a lens of fused silica. The214-nm channel with the glass lens is 214G and the onewith the Suprasil lens 214Q.

The diffusors are positioned in such a way that as therocket is spinning around its axis, the diffusors becomeexposed to sunlight one by one. The rocket also has aprecession, which of course influences the irradianceupon the diffusors (Fig. 2). To monitor this influence,the 450-nm channel is used, since in this wavelength

2528 APPLIED OPTICS / Vol. 20, No. 14 / 15 July 1981

214 G

450 < 4 .I 214Q

310 < - '256

I 1

288Mirors

D Diffusor: Lens4 Ni S04 -crystal

InterferenceI fitter

] Interferencemirror

- +-{G-l4 ._ --i-F-+_Photomultiplier

Fig. 1. Solar irradiance photometer.

-, I\

Sun

Fig. 2. Geometry of the occultation method. Also shown are themovements of the rocket (rotation and precession).

band extinction is negligible in the upper stratosphere.Figure 3 shows the spectral response of all channels.These spectral curves take the following into account:(1) the Fabry-Perot filter; (2) the NiSO4 crystal, if any;(3) the selective mirror, if any; (4) the PMT response;and (5) the solar spectrum.

The 214 curve splits into two branches above -250nm. The upper curve, 214 (Q-G), shows the effectivespectral response when the 214G channel has beensubtracted from the 214Q channel. Note that in spiteof the glass-quartz chopping, the 214-nm channel hasthe lowest selectivity of all channels. The 450-nmchannel has the smoothest and most selective spectralresponse, although this is the only channel that hasneither a NiSO4 crystal nor a selective mirror to improveselectivity.

The spectral response of the NiSO4 crystal and theglass diffusor are also shown in Fig. 3. The rapid dropof the 214Q curve above 330 nm is mainly due to ab-sorption in the NiSO4 crystal. Around 480-500 nm thiscrystal has another transmission window, and it is thiswindow that passes most of the unwanted light in the214 channel. This contribution is efficiently removedby the glass-quartz chopping technique.

Ill. Experiment

An opportunity of testing the glass-quartz choppingtechnique was given by the launching of two similarpayloads on 31 July (C59A) and 1 Aug. (C59B) 1971.The rockets were launched from Esrange, Kiruna.

10 -

200

Ni 504crystal

300 400 500 m

450iss lens

I

I I

t t At jcrysns

-0)

.3

04

Fig. 3. Spectral responses of all SIP channels and of the glass lensand the NiSO4 crystal. All responses are normalized to unity at theirpeaks. Note the two branches of the 214-nm response (See text for

details).

Table 1. Optical Components and Their Manufacturers in the Different SIP Channels

Channel 1 2 3 4 5 6Wavelength (nm) 214G 214Q 256 288 310 450Lens Glass FK-5 Suprasil I Suprasil I Suprasil I Suprasil I Glass FK-5NiSO 4 -6H 2 0 crystal Yes Yes Yes Yes Yes NoSelective mirror No No Yes Yes Yes No

Manufacturers of the optical elementsSuprasil I diffusors and lenses: Hereaus, BRDNiSO4 * 6H2 0 crystal: Semi-Elements, Inc., U.S.A.Interference filters: LKB-produkter, SwedenSelective mirrors: Schott, Mainz, BRDPhotomultiplier tube: EMI 9734-BQ

15 July 1981 / Vol. 20, No. 14 / APPLIED OPTICS 2529

214G

56 58 60 62 64 66 6 72 74 76 78 80 82 84 86 888 AMlttude km

214Q

56 58 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88Altitude km

Fig. 4. Uncorrected output from the 214Q and 214G channels ofC59B.

Each payload was equipped with one SIP, as previouslydescribed, but also with other instruments. These in-struments and the immediate results of all the experi-ments have been described by Witt et al. I

The purpose of the 256-, 288-, and 310-nm channelswas to measure the ozone absorption in different alti-tude ranges and to derive ozone densities. The originalpurpose of including a 214-nm channel was to monitorthe6 extinction of sunlight exciting the NO -y-band flu-orescence, which was measured by a different instru-ment in the same payload. This extinction can affectthe NO fluorescence at such low altitudes as 80-90 km.However, at the solar depression during this experi-ment, ,-2' or less, this extinction was negligible. Alsothe experiment took place in a strong display of nocti-lucent clouds, which made it impossible to measure theNO fluorescence below 90 km. Therefore, we used the214-nm channel to measure the ozone absorption in-stead and to compare the results with the other ozonechannels to find out how well the glass-quartz choppingtechnique worked.

The rockets were not altitude controlled but werespin stabilized with a spin rate of e-.7 rps, and the onlyattitude change was a fairly slow precession (Fig. 2).The output from the single EMI PMT was transmittedto the ground as a train of consecutive pulses corre-sponding to the six wavelength channels. During dataprocessing the channels are separated, sampled at reg-ular intervals, and then plotted. The plots for the 214G,214Q, and 450-nm channels of C59B are shown in Figs.4 and 5.

By plotting the envelopes of the dots in each plot, youwill get the maximum irradiance during each spin of therocket, corresponding to the minimum aspect angle

450

56 58 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88Altitude km

Fig. 5. Uncorrected output from 450-nm channel of C59B.

Fig. 6. Smoothed output from the 214Q and 214G channels (thegeometric calibration from the 450-nm channel taken into account).

Vertical scale in arbitrary units.

during the rotation, thus removing the effect of therocket's rotation. But this minimum aspect angle stillchanges slowly due to the rocket's precession, and thischange is monitored by the 450-nm channel (Fig. 5). Bydividing the outputs of the other channels with theoutput of the 450-nm channel, you will also remove theeffect of the rocket's precession and get the relative ir-radiance for each channel, as absorbed by the earth'satmosphere.

The relative irradiance of the 214Q and 214G chan-nels of both C59A and C59B have been drawn in Fig. 6.(This figure has been hand-smoothed and was drawnonly for illustration purposes. It has not been used inthe analysis.) Also shown in the same figure is thedifference between 214Q and 214G. The 214G channelmeasures only the unwanted light above 300 nm due topoor wavelength discrimination in the interference fil-ter. One can see that this stray light amounts to aconsiderable fraction of the total signal. At about a

2530 APPLIED OPTICS / Vol. 20, No. 14 / 15 July 1981

C59A- 214 nm

----- 256nm288nm

to measurement errors and inconsistencies in the datareduction such as the use of a standard atmosphere in-stead of the real conditions.

The solar spectrum needed for data reduction wastaken from Ackerman2 and Simon3 and the 02 and 03absorption from Ackerman. 2 We also used the CIRA1972 model atmosphere, 6 when correcting-for Rayleighscattering and 02 absorption. The numerical data re-duction technique has been described in more detailelsewhere. 4

IV. Discussion

Above -70-km altitude, the column densities derived5 60 6 70 75 80km from the 214- and 268-nm channels deviate from the

column densities derived from the 256-nm signal (Figs.Fig. 7. Intercomparison of ozone column density toward the sun, 7 and 8). The reason is that the column density, and

derived from the 214-, 256-, and 288-nm channels of C59A. hence the optical depth due to ozone absorption, aresmall, and thus the errors in the measurements becomelarge. At -250 nm, ozone has its maximum absorption,and therefore the 256-nm channel is the most reliableat high altitudes.

Between -60- and 70-km altitude the 214-, 256-, and-s 214nm288-nm channels agree quite well. However, below -60

10C 214nm km they again disagree. Here the 288-nm channel is the- -~~~~~~~~~~ ~ C5~ M256nm most reliable, since the optical depth at 256 nm is so

- 2nm large that the 256-nm signal becomes very noisy. In the10'7 " ~K<.... -. 214-nm channel, most of the signals at these altitudes

lo are due to the unwanted near UV visible light, which onthe other hand is monitored by the 214G channel.

I The 214-nm channel is not only affected by ozone1016 absorption. Actually the major part of the extinction

i-- - - is due to Rayleigh scattering and absorption by molec-ular oxygen. The other channels, 256 and 288 nm, arenot affected by molecular oxygen absorption or appre-

i15 ciably by Rayleigh scattering (see Table II, which shows55 sb 6 7b 75 Bbkm the optical depth at 60-km altitude, as observed by

C59A, and the relative contributions to this opticalig. 8d Intfrcompahriso 14of 2o5z6 ne cl2u88n densties to ward5hBsun, depth by ozone and oxygen absorption and by Rayleigh

derived frote14,56,nd288nmchnnlsofscattering).

Considering that a major part of the 214-nm signalis unwanted light and also that most of the extinctionof the measured light is due to absorption and scattering

30-km altitude, where the UV light below 300 nm is to of other gases than ozone, the agreement between thei large extent absorbed by the atmosphere, almost the 214-nm channel and the other channels must be con-entire signal consists of parasitic light, while at higher sidered quite good.altitudes where the atmosphere's absorption is negli- In Table II, it can also be seen that the optical depthgible, the unwanted light is still at least 25% of the total of the 310-nm channel, due to either ozone or Rayleighsignal. Using the glass-quartz chopping technique, one scattering, is very small. The 310-nm channel was in-,an easily eliminate the contribution of unwanted light tended to measure ozone at lower altitudes, where theibove 300 nm by a simple subtraction. ozone density and the optical depth of the other chan-

At the rocket's peak altitude the atmosphere's ab- nels become too large. Therefore, the 310-nm channelsorntion was neolio'ih1e in all wavelenoth chnnlk Thi is not included in this discussion.provides a calibration of the relative irradiances mea-sured by the instrument. Using this calibration, onecan derive the ozone column densities, i.e., the numberof ozone molecules in a column of air from the rocket,toward the sun and up to the top of the atmosphere.The column densities as derived from the 214-, 256-, and288-nm channels are shown in Figs. 7 (C59A) and 8(C59B). Ideally the column densities from the differentchannels should be the same, the differences being due

Table II. Optical Depth

Optical depth ai observed by C59A at 60 km due to:Ozone Oxygen Rayleigh

Total absorption absorption scattering214 nm 1.15 0.39 0.67 0.085256 nm 4.02 3.98 0.000 0.041288 nm 0.67 0.64 0.000 0.026310 nm 0.082 0.063 0.000 0.019

15 July 1981 / Vol. 20, No. 14 / APPLIED OPTICS 2531

I

cm-l

I

I

II

fIr

kI

that C59B actually measured ozone densities a little bitfarther north than C59A. Farther north the nights are

v- _- - c59A shorter during the summer; i.e., more sunlight is falling""^+ ------ C 59B onto the atmosphere, and one might speculate that more

10 ozone will be destroyed by sunlight farther north andthat this effect could eventually explain part of thedifference in measured ozone densities between C59Aand C59B. However, this effect must be very small,since the difference in solar depression, 0.70, whichequals the difference of the distance from the rocket tothe tangent point (which also can be expressed in de-

- grees), is quite small. Since the sun was in the north,* this corresponds to a latitude difference of only 0.70. If

we assume that the difference is due to a latitude effect,

50 55 60 65 70 75km the latitudinal gradient of the ozone density must havea very large value, far too large to be accepted. If we

Fig. 9. Computed local ozone densities C59A and C59B. make another assumption, namely that the differencein measured ozone density between C59A and C59B ismostly due to errors of 0.7 km in the altitude of one ofthe rockets is enough to explain the difference. If we

It should be pointed out that the instrument was make such an adjustment to the altitude of one of theIstructed and used in a time when fast concave rockets, the two ozone density curves will match toitings were not available. With the advent of the within "-5%. Therefore, we conclude that the sys-

lographic technique, multiwavelength occultation tematic difference between the two ozone density curvesloraphican tecnermuctiwved engh occungltan is due to errors in the measured altitudes of one or both,truments can be constructed by using a single entry rockets.illuminated by a diffusor and with detectors placed The compound error of the measurement causes the

the Rowland circle. The diffraction grating offers lack of overlapping between the individual curves inkarrower and more well-defined spectral transmission Figs. 7-9. From this the magnitude of the error can bein the Fabry-Perot filter, and occultation measure- es and, th he n error canents can readily be extended down to 170 nm. Still estimated, and, therefore, no error bars have been drawn-A f VA .- ... --- s -,, - ..4-. 11I-+;l in these figures.LIe uillwiuvely 6cLaUreU 11g16 uli ule grablig Will 6t'ili

make additional high-pass filtering necessary, i.e., bythe presently described technique. Also the space re-quired by a multiwavelength grating instrument maynot be compatible with the dimensions of meterologicalrockets.

Figure 9 shows the computed local ozone densitiesfrom both C59A and C59B. These local densities werecomputed by inverting the column densities and at eachaltitude choosing the column density corresponding tothe most reliable wavelength channel or, if severalchannels were about equally reliable, choosing the av-erage of these column densities. One can see that thelocal ozone densities are generally lower for C59B thanfor C59A. The difference is -17% on the average andhas the same sign throughout the altitude range. Partof this is due to measurement errors, since one cannotexpect a reproducibility much better than -10% usingthis method. In addition it is necessary to account forexpected uncertainty in the rocket altitude. However,part of the difference could also be due to a latitudedifference effect, since with the occultation method, asthe sun is below the horizon, the measurement is heavilyweighted by points close to the tangent point of the raybetween the rocket and the sun, i.e., the part of this raythat passes through the lowest altitude in the atmo-sphere (Fig. 1). The tangent points are often several100 km away from the rocket.

During the C59A flight, the solar depression was 1.40,while it was 2.10 during the C59B flight. The sun wasin both cases approximately in the north. This means

V. Comparison with Other Ozone Measurements

When surveying the literature, we have not been ableto find any other ozone measurements during a twilightsituation at the same altitude, latitude, and season.Therefore, we can only compare with our other SIPflights. Since the main purpose of this paper is to de-scribe the performance of the glass-quartz choppingtechnique, we will only briefly compare the differentflights.

There has been no SIP except that in C59A and C59Bequipped with a 214-nm channel. Several other SIPshave had channels in the ozone absorption bands,however. Two SIPs were launched in the summer of1969 and 1970 from Esrange. They both yielded ozoneconcentrations not significantly different from the onespresented here. These data are not yet published.

On 30 July 1978, a concave grating instrument formeasuring solar irradiance was launched from Esrange.This instrument had a 250-nm channel which yieldedozone concentrations very close to the C59 results. Ifdrawn in Fig. 9 this curve would fall between the onesof C59A and C59B. The data are presently being pro-cessed.

On 13 Mar. 1975, another SIP was launched fromEsrange. 5 Above -55 km the ozone concentration wassignificantly lower during this flight. However, thisflight was during a different season.

The authors with to acknowledge the assistance ofNathan Wilhelm for building the instrument and to

10

10I

co]grzhoi nsslilona rth,mcME,

2532 APPLIED OPTICS / Vol. 20, No. 14 / 15 July 1981

thank Margareta Gustafsson for drawing the figures.This work was supported by grants RN/1-39 andDR2179 from the Swedish Board of Space Activities.

References1. G. Witt et al., Sounding Rocket Experiment for the Investigation

of High-Latitude Summer Atmospheric Conditions Between 60and 110 km (ESRO Experiment R-185, Payload C-59), ReportAP-7, Department of Meteorology, U. Stockholm (Dec. 1971).

2. M. Ackerman, "Ultraviolet Solar Radiation Related to MesosphericProcesses," Aeronom. Acta 77, xxx (1970).

3. P. Simon, "Nouvelles measures de l'ultraviolet solaire dans lastratosphere," Aeronom. Acta 145, xxx (1975).

4. S. Grahn, and G. Witt, "High-Latitude Ozone Soundings with aRocket-Borne Multiwavelength Solar Irradiance Photometer,"Report AP-16, Department of Meteorology, U. Stockholm (Sept.1974).

5. E. J. Llewellyn and G. Witt, Planet. Space Sci. 25, 165 (1977).6. Cospar International References Atmosphere 1972 (Akademie-

Verlag, Berlin, 1972).

TRANSFER OF COPYRIGHT AGREEMENT

Title of article:

Author(s):

The undersigned hereby assign(s) to the Optical Socity of America (the "Society") copyright ownership in the above-titledarticle, this assignment to be effective as of the date of acceptance of this article for publication in the above-named journaland to the extent transferable under applicable national law.

Authors (or their employers, in the case of works made for hire) reserve the following rights:

(a) The right to use all or part of this article in future works of their own, including, but not limited to, lectures, lecturenotes, press releases, reviews, textbooks, or reprint books;

(b) The right to direct the Society to grant permission to third parties to republish all or part of the article or trans-lations thereof, provided, however, that without any limitation whatever on such right, the Society shall be entitled,at its option, to negotiate and to receive reasonable royalties in respect of any such republication;

(c) The right to refuse permission to third parties to republish all or part of the article or translations thereof, with theexception of the permission contained in the copyright statement appearing in each issue of the journal.

The Society shall retain all proprietary rights with respect to journal issues as a whole.

Signature

Print name

Title

Signature

Print name

Date Title

Signature

Print name

Date Title Date

Who Should Sign

This form must be signed by(1) each author who is not a U.S. Government employee, if reasonably practicable; and

(2) each such author's employer, if the article was prepared by such authorls) within the scope of his or her employ-ment or was specifically ordered or commissioned by the employer.

If any author was a U.S. Government employee at the time this article was prepared, and such article was prepared as part ofhis official duties, such author should sign the following form:

I hereby certify that this article was prepared by me as part of my official duties as an employee of theUnited States Government.

Signature

Print name

Title

Signature

Print name

Date Title

Signature

Print name

Date Title Date

This signed statement must be received by the editor's office before the article can be published. Address requests for further in-formation to the Executive Director, Optical Society of America, 1816 Jefferson Place, N.W., Washington, D.C. 20036 (telephone:202 223-8130).

15 July 1981 / Vol. 20, No. 14 / APPLIED OPTICS 2533


Recommended