OXIDATIVE DEALKYLATION OF A LKY LA RO MA T I C HYD ROCA R BO NS Selective Vapor-Phase CataQtic Oxidation of Alkylaromatic Hydrocarbons to Their Parent Homologs C H A R L E S J . NORTON A N D T H U R L E E. MOSS' Denver Research Center, .\iarathon Oil Co., Littleton, Cola.
Alkylaromatic hydrocarbons may b e selectively dealkylated to lower homologs under conditions of controlled vapor-phase catalytic oxidation. Various metal oxides were screened as catalysts. Cadmia proved effective. Effects of temperature, oxygen concentration, and cadmia concentration were studied using a gas-liquid chromatographic microreactor and an experimental design. Yields and conversions were correlated with variables b y least squares fitting of the data to second-degree polynomials. Best results for the oxidation of 1 -methylnaphthalene to naphthalene were obtained a t 400" to 450" C. over 2 to 8 weight cadmia with 5 to 25 weight % oxygen, yielding 30 to 62 i 5 weight naphthalene at conversion levels of 20 to 25 f 3 weight
70
investigations of the catalytic oxidation of alkylnaphthalenes to phthalic anhydride over vanadia catalyst, we discovered significant amounts of naphthalene in the recovered reaction product mixture (6). Stein et al. (70), in microreactor investigations of catalysts for complete combustion of pure hydrocarbons, observed significant differences in the ease of oxidation of various hydrocarbons. O u r preliminary experimental rrsults, together with interpretation of Stein's data. suggested that it might be feasible to oxidize the alpha-positions of alkylaromatic hydrocarbons selectively in preference to the aromatic ring carbon-hydrogen bonds. If the oxidation proceeded further to form a carboxylic acid group, this group might decarboxylate under the reaction conditions and provide a novel process for the oxidative dealkylation of alkylarornatic hydrocarbons to their more commercially valuable parent homologs: URISG
H I
,
%
Yo.
$naphthalene, 2-methylnaphthalene, 1,2-, 1,6-, and 1,7dimethylnaphthalenes, and ,a light catalytic cycle oil extract rich in polymethylnaphthaleries. Experimental
Reagents. H'iDROCARBOIU. FEEDS. Several hydrocarbon feeds were studied extensively for selective oxidation. Most of the screenin.: \vas done on a mixture of 25 weight % naph' 1-methylnaphthalene in order to thalene and 75 weight 7 confirm the naphthalene retention time under the reaction conditions. Several runs were made with pure 1,2-, 1,6-, and 1,7-dimethylnaphthale:nes3 toluene, xylenes, and an aromatic extract of light catalytic cycle oil. The reagents used Lvere reagent-grade or better. Available m.eta1 oxides, most of which were C.AT.ALYSTS. solid and stable u p to a temperature of 550" C . under the reaction conditions, were cltudied, but volatile oxides and alkali and alkaline earth oxides (which would react with carbon dioxide) were omitted (Figure 2). CATALYST SUPPORTS.Materials investigated as supports are listed in Table I.
catalyst
R Table 1. L
J
L
-I
\Ve decided to investigate the practicability of this oxidative dealkylation reaction in comparison with other dealkylation reactions (2). J f i t h commercial vanadia catalyst, yields and conversions of 1-methylnaphthalene to naphthalene were limited and lo\v ( 7 ) . so a n extensive screening program was undertaken to find a more selective catalyst for this purpose. A microreactor technique, combined with direct gas-liquid chromatographic analysis (4.70), was used to investigate and compare a number of pure, reagent-grade metal oxides and other forms of catalysts for their selective catalytic activities for the production of naphthalene. One of the more effective catalysts, cadmium oxide o n silica, \vas also investigated for the dealkylation of toluene, xylenes, Tetralin, Decalin. l-meth-
support
Catalyst Supports Investigated Performance Source at 400' C.
Aluminas Activated Alundum Silicas Davison 923 Microspheroidal Grade 12 Grade 13 Grade 70 (140 A. pore) Prepared Silica-aluminas 1 57 0 alumina 257' alumina Molecular Sieves Ceramic Silicon Carbide
Alcoa Norton
Adsorbent Nonadsorbent
Davison Davison Davison Davison Davison Pptd. with "OB
Adsorbent Adsorbent Adsorbent Adsorbent Adsorbent Nonadsorbent
Davison Davison Linde Insulation beads Carborundum
Adsorbent Adsorbent Adsorbent Nonadsorbent Nonadsorbent
Present address, LMarathonOil Co., Findlay, Ohio. VOL. 3
NO. 1
JANUARY 1964
23
Apparatus. A gas-liquid chromatograph designed and built a t the Denver Research Center of the Marathon Oil Co. was used. This unit has a second thermostated oven in which precut columns, gas samplers, and other additions to the apparatus may be installed or removed while keeping the connecting tubing hot. Compressed gas cylinders of air, oxygen, nitrogen, and airnitrogen blends were used as sweep gases. 4 1-mv. recorder was used because the sensitivity of the thermal conductivity cell was low when air was used as a siveep gas. Silicone rubber sampler gaskets rapidly deteriorated, but custom-made Teflon gaskets performed well. The microreactor heater consisted of a 35-cm. section of 3/8-inch i.d. copper tubing, wrapped with a layer of 0.010- X 1.5-inch glass tape. Thermocouples were taped to the center of the wrapped tube, and the assembly was wrapped with resistance wire and glass insulation tape. Thermocouple leads were connected to a Brown indicating controller and/or a Minimite temperature indicator lvhich had been calibrated. Each microreactor tube was constructed of a 40-cm. length of 0.035-inch wall, 1/4-inch stainless steel tubing. After being filled with catalyst, the tube was inserted into the heater and connected in line with Swagelok fittings in the left oven. A fresh reactor tube was made for each catalyst. Sample injections were effected with custom-made brass microdippers. A 0.07-ml. microdipper in a Teflon selflubricating sleeve holder was found to deliver a reproducible feed sample of suitable volume for most of our studies. Procedure. CATALYST PREPARATION. For screening purposes, 0.4 gram (0.5-ml. volume) of -100- +200-mesh, dry metal oxide was weighed on an analytical balance and mechanically dispersed in a vial onto 6.0 ml. of dry -30f60-mesh Alundum which had been ignited at 600 C. Procedures tested for the preparation of cadmium oxide on supports and their performances are summarized in Table 11. Method 7 yielded the most active catalyst preparation for later investigations. Samples of cadmia on amorphous silica were prepared by this method a t three levels of cadmium con-
Table
II.
Various
.Wethod 1. CdO slug-Alundum
Preparations of Investigated Form
Cadmium
Catalyst
Performance
CdO
Good at high tempera-
CdO
Better at high temperatures than 1 More active than dispersion 2
tures
2. CdO dispersed on
Alundum 3. Evaporation of aqueous
CdT';Os on Xlundum followed by ignition Vacuum impregnation of aqueous C d N 0 3 on Alundum followed by ignition Decomposition of Cd(OH)2 on Alundum Decomposition of CdC03 Coprecipitation of Cd f2 with silica ; ignited at 600" C.
4.
5. 6. 7.
CdO CdO
Better than arm. evaporation, technique 3
CdO
Similar to 3
CdO
Similar to 2 and 3
Amorphous CdO/SiO?
Most active and selective catalyst at 400" C.
Table 111.
a
24
Analyses of Coprecipitated Cadmia on Silica Catalysts C d O Found, 70 Chemical X-ray, h'ominal wt. 70CdOa analysis analysis 2.5 1.20 1.88 5.0 2.95 3.43 7.5 6.08 7.88 Calculated from amount of cd(,vo3)z. 4 HzO used.
I&EC PROCESS DESIGN A N D DEVELOPMENT
(4)
-
Separation Column
Exhoost
Detector
f9oi
Air a t ' l 7 p.s.i.g.
L
Air
ow ntrol rlves
(9)
Rotumetc .a.,.y.
I
Ai1r Cylinder
Figure 1.
Schematic diagram of microreactor apparatus
tent. Amounts of cadmium nitrate were chosen to give the theoretically nominal percentages of cadmium oxide equivalent: 2.5, 5.5, and 7.5 weight % CdO. The resultant catalyst preparations were analyzed by x-ray and wet chemical techniques (Table 111). T o a solution of 45 ml. of N brand water glass in 855 ml. o distilled water was added dropwise a second solution of 2.4 grams of Cd(N03)2.4 HzO in 100 ml. of distilled water over about 10 minutes to produce a milky suspension. Four drops of phenolphthalein were added, and a third solution of aqueous 10 volume % nitric acid was added dropwise with vigorous stirring over a period of about 10 minutes, carefully dispelling the pink color to produce a final p H of 6, as determined by indicator paper. The resultant solution was poured into centrifuge bottles, allowed to set for I/z hour. and centrifuged. The supernatant solution was decanted. The gel was mixed and rinsed five times with portions of aqueous 10 weight yo ammonium nitrate solution. It was then dried 16 hours at 138' C., ignited at 600' C. for 2 hours in a muffle furnace, weighed, mixed, pulverized, and screened to - 30- +60-mesh size. A sample of this catalyst was shown by x-ray fluorescence analysis to contain 3.43 weight % cadmium. Diffraction x-ray analysis indicated the material to be amorphous in structure. -4fter heat-treating a t 850' C . for 4 hours, crystallites of sufficient size were grown to be identified as silicon dioxide and cadmium oxide. Used catalyst also exhibited these characteristics. MICROREACTOR INSTALLATION AND OPERATIOS. A cleanf microreactor tube was plugged a t one end with a small wad of borosilicate glass wool; 6.5 ml. of -30- f60-mesh catalyst was added and packed with the aid of vibration. The other end of the microreactor was plugged with glass \vool. The packed microreactor was slipped through the heater, equipped with Swagelok fittings and 0 rings, and sealed into the reactor line system in the left oven of the gas-liquid chromatographic apparatus. All of the apparatus except the air cylinder \vas contained in t\vo air baths of the gas-liquid chromatographic apparatus thermostated a t 200' =t5' C. (Figure 1). The microreactor tube, 1, was inserted into the resistance heater, 2. which was maintained at a preset reaction temperature in the range 300' to 600' C. Bypass valves, 6, could be set to pass the feed in the air stream through the microreactor or to bypass the microreactor. A pressure differential of about 4 p.s.i.g. across air lines 7 and 8 was maintained at a fixed rotameter flow setting, 9, by flow control valves, 10. a t the sample injection block, 11. Hydrocarbon samples of 0.001 to 0,010 ml. were injected with calibrated microdippers. The drying column, 3, was packed \vith a 50 : 50 mixture of Ascarite and magnesium perchlorate to
Table IV. Retention Times of Reference Compounds under Gas-liquid Chromatograph-Microreactor Conditions Retention Time,a M i n . 2.5 co 2.5 co:! Benzeneb 3.0. 3.5 3.4. 3 7 Tolueneb
Maleic anhydride H20 m-X\,lene* p-xylene" Citraconic anhydride Saphthalene 2-Methylnaphthalene Phthalic anhydride 1-Methylnaphthalene 1,4-DihVdroxynaphthalene 1:4-Naphthoquinone 1-Saphthol 2-Methyl-l,4-naphthoquinone 1-Naphthaldehyde 1-Naphthoic acid
1,2-Naphthoquinone
35,42 3 7 38,43 3.8, 4 . 2 4.25 8.2 10.0 10.6 12.2 13.8 14.4 18.0 18.2
21.1 33 50
Standard deointion i n rrtention times is about &0.8 minute. C'nder these conditions, siwaral pure, lowboiling compounds give double peaks.
remove most of the carbon dioxide, water, and maleic and phthalic anhydrides which interfered with the hydrocarbon analysis. The separation column, 4, was 2 meters in length, of 0.25-inch 0.d. stainless steel, with C column packing of 1.8 grams of Chromosorb 154-0252. ANALYSIS.Retention times of pure feed materials, reaction intermediates, and reaction products are summarized in Table IV. Reproducibility of the retention times, as well as the peak areas: \vas ensured by using a standard, narrow catalyst particle size range (-30 f6O mesh) to control the pressure drop in the microreactor. Peak areas \vere measured by a planimeter. An average of three or four area measurements was taken to minimize the error of this operation to less than 1 1 % . Calibrations with the feed mixtures were obtained by bypassing the microreactor. The gas-liquid chromatograms so obtained gave an analysis of the feed mixture and a measure of the total feed. Feed mixture calibrations were run a t the beginning, in the middle, and a t the end of a series of runs each day to get a good average calibration and to minimize the effect of any experimental drift during the series of runs. Calculations. REACTION COSDITIONS.Reactor length, catalyst volume, flow, and sample size were maintained constant for all the screening runs, so that air-to-feed ratio and contact time were essentially constant. Only the temperature of the reaction was varied over the range 300" to 600" C. (The catalyst bed temperature was checked and found to be the same as the heater jacket temperature under flow conditions.) T h e air-to-feed ratio was estimated to be 0.2 to 1.3 liters of air per gram of hydrocarbon feed. The air-to-feed ratio was varied in the studies over cadmia by varying the oxygen content of the carrier gas stream. Per cent combustion, composition of recovered hydrocarbons, and naphthalene production, conversion, and yield were calculated by the equations:
yGHydrocarbon recovery
Wt. Wt.
70
= 100% -
x
100%
naphthalene yield = amount of naphthalene produced (wt. 70 1-methylnaphthalene in feed X total peak area) (wt. 7 0 1-methylnaphthalene in product X total product peak areas) 70
(8)
PRECISION.Analysis by gas-liquid chromatography is o n compositions in the range of accurate to about =t0.5yG 10 to 90%. Statistical analysis of hydrocarbon recovery data obtained for a number of runs a t the same conditions showed a deviation of =t5.0% or less. Because of the mathematical steps involved in the calculation of yield, the propagated error in the calculated yield may be much larger than either of these values. At conversion levels below 15%, uncertainty in I-methylnaphthalene consumption amounts to almost 507, of the value obtained; a t moderate conversion levels. it is less than 10% of the value obtained. Actually, the yields obtained a t even the low levels of conversion are probably more accurate than rigorous calculation of precision would indicate. The major source of error is the hydrocarbon recovery, estimated a t 1 5 % , which may be too large. The data points are reproducible a t all levels of conversions and yields investigated and are accommodated nicely by a smooth curve. The variance, or lack of fit, obtained from cadmia studies is considerably less than this estimation of error. Results and Discussion
Oxidation of 1-Methylnaphthalene to Naphthalene. SCREENING O F METALOXIDES. Metal oxides screened are summarized in Figure 2. Results obtained in a typical experimental run over cadmium oxide are described below. The microreactor tube was filled rvith 0.400 gram (0.5ml.) of - 100- +200-mesh cadmium oxide mechanically dispersed on 6.0 ml. of -30- +60-mesh Alundum support. The microreactor was brought to 525' C. A40.07-ml. sample of pure I-methylnaphthalene was injected into the air stream a t a n air flow of 17.5 ml. per minute a t standard temperature and pressure, bypassing the microreactor to obtain the gas-liquid chromatographic calibration curve, 1, of Figure 3.
IlIl I
I
1 +I 1+21 '
3
+3
1
14~5,6171 6
1
112]mlE/Yim~ZlIlo
VALENCES
I+ 31- 4+41-3+51- 2+el -I + I ! 0
VARIABLE
' 1 1 2
I
'
I
yo hydrocarbon recovery
(2)
(3)
naphthalene peak area naphthalene = total hydrocarbon peak areas
5% 1-methylnaphthalene
+ Figure 2.
=
1-methylnaphthalene peak area
__.
Wt.
=
recovered hydrocarbon peak areas calibration peak areas
% Combustion
Amount of naphthalene produced = wt. 70 naphthalene in product X total peak areas of product -wt. % naphthalene in original feed X total peak areas of feed (6) Wt. 7cconversion = amount of naphthalene produced total 1-methylnaphthalene feed x 100% (7)
total hydrocarbon peak areas
x
100% ( 5 )
90
91
92
93
94
95 196
97
98
99
1
oc
0
02'
i
Th Pa U Np PuiAm,Crn Bk Cf Es 5 n M d i N o l
Screening of metal oxides
Metal oxides screened as selective oxidation catalysts far naphthalene production indicated by heavy lettering
VOL. 3
NO. 1
JANUARY
1964
25
100
I
A
1
f
Ill TOTAL I-METHYL- NAPHTHALENE FEED
1 3 ) ~
80
_-
c
I-
870
o
i?
n
3
W
a
8 €0
[L
a Y 4. W
z
50
n
W
40
W
z W J
? I4. - 30
a
J
c
I
g 20
S
a
a
z
10
z 0
+ 3
2 6 6 W l % Yield 0
2
6
4
8
RETENTION
10
TIME,
12
14
16
I8
Catalysts
Selectivitv for Combustzon iV/7-M.V Mzxture
Production of from
7-MN
-4. Metallic oxides
Be0 .41r03
SiO,, TiO?,V?O,, Crz03 MnOp, Fen03, CoaO4, X i 0 CuO, ZnO, GazO3, GeOn, Y203,MOO? .4g?O CdO, In*Oi, SnO? SbrOs
+ ++
t + T e
Cadmium compounds
B.
CdF2 CdS CdSe ~_.. CdTe CdMoOd CdSiOa Cd ZBP W p 0 32
-
++ 6 + + E
+
Bi.0,
++ +f ++ +
+-+ ++ T
f
f
a z41undum, an alumina-silica, gave little combustion at any temperature and w a s used as a support ,for testing the other oxides. * Silver oxide decomposes at 300' C . but naphthalene w a s produced over the decomposition product, which w a s probably the metal. Antimony pentoxide decomposes to Sb?Oa at 380" C. and adsorbs all feed, yielding a new compound (unidenti'jfed) at 200" C. Complete combustion w a s noted at 300Q C. Platinum oxide gave complete combustion a t 200" C. e Mercuric oxide decomposes to metaliic mercury under reaction conditions, and the volatile metal is carried out of the reaction zone. f Little reaction occurrpd ouer cadmium telluride at 5.50' C.
26
20
WT
Catalytic Activities
l & E C P R O C E S S D E S I G N A N D DEVELOPMENT
I
1
50
60
I 10
MINUTES
Figure 3. Selective oxidation of 1 -methylnaphthalene to naphthalene over cadmia at 525" C.
Table V.
IO
Figure 4. tion series
30
40
% COMBUSTION
70
1
1
80
90
100
OF FEED
Selective oxidation in first transi-
The microreactor was then put in line, and a 0.0'-ml. sample of pure 1-methylnaphthalene was injected into the air stream at the same air flow rate of 17.5 ml. per minute. The estimated reaction conditions were 525" C., 0.2- to 2.0-second contact time, and 0.1 to 5.0 liters of air per gram offeed. The reaction product mixture gave the second chromatographic curve, consisting of carbon dioxide, 3, which went off-scale a t the top; water, 4, which went off-scale at the bottom; a naphthalene peak, 5 ; and unconsumed 1-methylnaphthalene, 6. The difference in areas between the total 1methylnaphthalene obtained on calibration, bypassing the microreactor, 1, and the area of the recovered 1-methylnaphthalene, 6, obtained on passing the hydrocarbon in the air stream through the microreactor catalyst bed a t 525" C., gave the 1-methvlnaphthalene consumed, 7. The conversion to naphthalene vias 24.0 weight yc per pass, based on the total 1-methvlnaphthalene feed. The naphthalene yield was 36.6 weight %, or 40.6 mole %, based on the actual amount of 1-methylnaphthalene consumed.
SELECTIVITY COMPARISONS. Most of the metal oxides investigated effected significant selective catalytic oxidation of the 25 weight % naphthalene-75 weight % l-methylnaphthalene feed mixture considerably below the spontaneous ignition temperature. Some catalysts are extremely selective in their preferential oxidation of 1-methylnaphthalene (Table V). Figures 4, 5, and 6 summarize and compare screening results with the various catalysts. Figure 4 plots the naphthalene content in the recovered naphthalene-1-methylnaphthalene mixtures a t increasing extents of combustion from 300' to 600" C. The original feed was 25 weight % naphthalene and 75 weight 1-methylnaphthalene, as indicated a t 0% combustion of the feed. If only the methjl group Lvere burned in the feed mixture, the recovered hydrocarbon Xrould approach 100% naphthalene a t about 10% combustion. In Figure 4 the stable metal oxides of the elements of the fourth row of the periodic chart, titanium through germanium, are compared ; vanadia has the most desirable selectivity-that is, produces the highest naphthalene enrichment in the recovered hydrocarbons a t the least degree of total feed combustion, Several of the first transition series metal oxides-e.g., chromia-exhibit good selectivity a t low per cent combustions of the feed but become less selective a t higher extents of combustion. Catalysts differ in their temperature coefficients of selectivity for the feed components.
IO
3 0 Q 0
a
T W
2
W
z
W
*tA
W _I
4
r
I-
t
L
n
I
a
q
2
4
z
8 __--
__
+ 3 0
10
20
WT
.50 '/o
40
50
60
COMBUSTION
70
OF
80
90
IO0
WT. W COMBUSTION OF F E E D
FEED
Figure 5. Oxidation selectivities of conventional catalysts for phthalic anhydride production
Figure 5 compares vanadia and other catalysts Lvhich hlarisic (5) investigated extensively. The order of activity reported by Marisic for phthalic anhydride production from naphthalene is 1 . 2 0 5 > MOOS> \vO3. Our results show a similar order of selectivity for the oxidation of 1-methylnaphthalene in competition with naphthalene. The stable oxides of cadmium, bismuth, and indium were as selective as or more selective than vanadia (see Figure 6). Cadmium oxide is the most selective oxidation catalyst discovrred in our screening program. it'ithin experimental error of about =7yc.the results obtained with the thvo feed mixtures are about the same. Naphthalene >-ieldsover C d O are in the range of 54.0 zz 6.6 iveight % and are considerably better than the yields attainable over the next best catalysts: indium oxide ( I n & 26.9 = 5.1 \\.eight 70): and bismuth oxide (Bi903, 24.1 rt 6.6 \\.eight 7,. :\ll of these catalysts are superior to a commercial fluid-vanadia catalyst, for Ivhich the best yield is 8.1 iveight The metal oxides listed in Table V lvere also investigated for the direct production of naphthalene (S)from pure l-methyinaphthalene (1-MK). The production of naphthalene (+ in column 3) was achieved a t 300' to 600' C., about 0.1- to 2.0second contact time. and about 0.1 to 5.0 liters per gram air-tofeed ratios-conditions considerably different from thermal or catalytic hydrocracking conditions. FURTHERSCREENINGOF CADMIUMCOMPOUNDS. Since cadmium oxide was one of the most selective oxides for naphthalene production, several other cadmium compounds prepared in the same fashion were tested (Figures 7 and 8). Cadmium sulfide, cadmium selenide, and cadmium telluride \Yere found to be catalytically active and to manifest oxidation selectivity, but were all less active and selective than cadmium oxide. Cadmium mol>-bdate,cadmium silicate, and cadmium borotungstate were moderately active and selective, but less than cadmium oxide. X-ray inspection of these catalysts after use showed all but cadmium telluride to be substantially unchanged. Cadmium catalysts were also prepared by a number of alternative methods (Table 11). The most active and selective catalyst was prepared by the coprecipitation of cadmium hydroxide Lvith silica gel. 'This catalyst, after drying and igni-
Figure 6.
Oxidation selectivities
Most selective oxidation catalysts are from right side of periodic chart
w
z
W -1
a
I L
8
a z
-
z,
IO
20
30
WT
% C O M B U S T I O N OF F E E D
40
50
60
70
80
90
100
Figure 7. Effect of simple anions on cadmium oxidation selectivity
100,
,
,
,
CdO
l
1
'
,
90 80
70
z 60
50 40
30
P 0
I0
20
WT
30
40
50
60
% COMBUSTION
70
OF
80
90
100
FEED
Figure 8. Effect of complex anions on cadmium oxidation selectivity VOL. 3
NO. 1
JANUARY
1964
27
tion, is opaque white. The original crystallites in this ignited catalyst were too small for detection of cadmium oxide. However, x-ray fluorescence spectroscopy can be used to assay its cadmium content. The cadmium oxide and silica crystallites grow under the reaction conditions and are detectable after the catalyst has been used. CATALYST SUPPORTS.In preliminary experiments we tried several materials as supports and found Alundum to be the most inert and least catalytic material available. Therefore, our screening experiments were carried out with this material
600'
C. 10
40
30
20
primarily as a filler in the microreactor. In later experiments, to observe the effect of the support o n the activity and selectivity of cadmium, we obtained more data on a number of materials that might be used as supports (Table I). Of the nonadsorptive supports, the specially prepared amorphous silica coprecipitated with cadmium appears to be the best. FEED MATERIALS.Table VI summarizes the qualitative results for the oxidation of various feed materials over cadmia. Lower homolog aromatics were obtained as reaction products from the aromatic hydrocarbons toluene through dimethyl-
70
60
100
80 60 40
20 0
* a
w 100
50OoC.
>
0 80 0 W
U
60
n 40 W
W LL
20
0
I-
z W 0 100
U W
80
a 60 I-
I 40
-
c3
20
W
3
0
100 80 60 40
20 0
A T O M I C N U M B E R OF M E T A L Figure 28
9.
ION
Electronic effects on catalytic activity
l & E C PROCESS D E S I G N A N D DEVELOPMENT
80
90
Feed Toluene Xylenes Tetralin Decalin 1-Methylnaphthalene
Oxidant
1-Saphthaldehyde
Air
1,2-Dimethylnaphthalene
Air
1070 0
1,7-Dimethylnaphthalene Air 1.6-Dimethvlna~hthalene Air LCCO extract (490'525' F.) ,
Table VI. Oxidation of Various Feeds Conditions Catalyst T R X ,O C.
Cadmium copptd. with silica Cadmium copptd. with silica Cadmium copptd. with silica Cadmium copptd. with silica Cadmium oxide or cadmium copptd. with silica
400-600 400-600 400-600 400-600 400-600
Reaction Products Benzene in small amounts Benzene, toluene in small amounts Naphthalene in good yields Naphthalene in trace amounts Naphthalene in good yields
Cadmium copptd. with silica
400-600
Naphthalene in small amounts
Cadmium copptd. with silica
580
Cadmium oxide dispersed in Alundum
442-580
NaDhthalene. methvlnaDhthalene. dimeth;lnaphthalene in 'small amounts
Cadmium oxide dispersed in Alundum
400-600
2
I
naphthalenes. The oxidative dehydrogenation of Tetralin and Decalin to naphthalene was also confirmed. Saphthalene and methylnaphthalenes were produced from a light catalytic cycle oil aromatic extract cut rich in dimethylnaphthalene isomers. Yields ranged from trace amounts to good yields, but are not indicated because of insufficient investigation. Several probable reaction intermediates in the oxidation sequence of I-methylnaphthalene to naphthalene were also injected as feeds. It was difficult to get enough I-naphthyl alcohol or I-naphthoic acid into solution for injection, and in neither case was naphthalene detected as a product. 1-Naphthaldehyde did give naphthalene as a reaction product. Review of Stein's (70) oxidation studies a t 80% combustion levels showed a n apparent electronic effect in his results and a sharp change in the oxidation specificities of heptane as compared with benzene between oxides of copper and zinc. This distinction was not so dramatic in our oxidation studies, Lvhich \rere carried out a t lo\v combustion levels, but this comparison focused our attention on cadmium and the adjacent region of the periodic chart. A definite nonrandom electronic effect is apparent in our screening results \Then hydrocarbon recoveries are plotted against atomic numbers of the metals of the oxides a t various reaction temperatures (Figure 9). These variations in catalytic activity of the cation are different from variations in the electronegativities of the metals (8). O n the other hand, there may be a correlation between the anion electronegativity end catalyst activity (Figure IO). Correlation of Experimental Results. Experimental design and statistical methods for investigating chemical reactions for their optimum conditions, described by Davies ( I ) : applied by Franklin ( 3 ) , and mechanistically interpreted by Pinchbeck ( 9 ) . were used to study and correlate the effects of the major selective oxidative reaction variables upon naphthalene yields and conversions from 1-methylnaphthalene over cadmia-silica catalyst. The variables chosen for our gasliquid chromatographic-microreactor studies are shown below to be linearly related to those studied by Franklin ( 3 ) for optimum phthalic anhydride production from naphthalene over vanadia in air. .4n experimental design embracing four levels of temperature (400', 450°, 500". and 550' C.), three levels of cadmia o n silica (1.18, 3.43: and 7.88 weight % CdO, by x-ray analysis), and three levels of oxygen (5.0, 12.0, and 20.0 weight y6 oxygen) was completed to determine the effects and interactions of these variable factors upon naphthalene yields and conversions (Figures 11: 12, 13: and 14 and T a b l e V I I ) .
Naphthalene, methylnaphthalene, dimethylnaphthalene in small amounts
The superficial contact time in the microreactor can only be estimated from the gas flow rate and catalyst volume. This value was estimated to be 0.2 to 2.0 seconds. Ho\rever, the effective contact time can be varied by altering the concentration of the active cadmium sites o n the fixed catalyst surface. If we assume the cadmia is well dispersed o n the support a t the various concentration levels, then the effective contact time is proportional to the cadmia concentration : t = a
(wt. yo CdO)
(9)
Taking the logarithm of both sides of this equation gives log t
=
log (wt. % CdO)
+ log a
(10)
Therefore, log t is linearly related to log (irt. % C d O ) , differing only by an unknown constant?log a. Similarly, it can be demonstrated that, with the fixed injection feed volume, varying the oxygen concentration in the cnrrier gas stream is equivalent to changing the oxygen-to-feed ratio : Os/F
=
b (wt.
70 0 2 )
(11)
or log Oa/F = log (wt. % 0 2 )
+ log b
(12)
Therefore, log Oz/F is linearly related to the log (Jvt. 76 0 2 ) . differing only by a n unknoirn constant, log b. Experimental results obtained a t each temperature level were treated as isothermal groups 2nd fitted by least squares procedures to a second-degree polynomial equation of the general form a
=
4 .
+ Rx + cy + Dx* + e\'?+ F.uy
113)
where cy is the weight naphthalene yield. I is the log (wt.%, CdO), and y is the log (wt. % 0 2 ) . Calculations \vere carried out o n an Electrodata Datatron 205 digital computer. Coefficients for the best fit of all the data points of each group of data were obtained. The fact that we have more than six data points in each group permits the calculation of a n experimental error, variance, o r lack of fit? for each group of data. The variance for each group of yield data becomes less at higher temperatures. An observed contributing factor to experimental error is physical adsorption, even a t 350' to 400' C., and this is greatly reduced a t higher temperature. In general. the data fits are good and within the estimated error discussed under "Precision." The isoyield curves a t 450' C. bvere shifted somewhat by omission of one or two yields, but the character did not change much. VOL. 3
NO. 1
JANUARY 1964
29
~~~
Table VII.
Data Point
.vo.
Key to Isoyield-lroconversion Surfaces on Figures 1 1 to 14 Nabhthalene Results Yield, Wt. yo Conversion, Wt. % log [% 0 2 1 Exptl. Predicted Exptl. Predicted ['%??do] %02 0.274 0,699 22.0 19.9 5 .O 4.7 5.3 0.535 0.699 5 .O ... ... ... ... 0,897 55.0 5 .O 4.1 52.9 4.7 0.699 0.274 51.9 50.0 16.4 12.0 15.2 1.079 59.5 0.535 26.3 52.9 12.0 31.2 1.079 0.897 18.9 75 .o 12.0 15.2 1.079 66.5 0.274 12.4 20.0 38.6 38.8 14.2 1.301 0,535 23.2 48 . O 41.4 20.0 18.3 1.301 41.5 20.0 20.0 0.897 1.301 35.2 16.9 f5.4 Error 1 3 . 1 0.274 0,699 12.5 10.9 72.9 69 .O 5 .O 55.3 13.8 58.9 0.535 0.699 10.1 5.0 32.6 0.897 32.3 0.699 5.3 5 .O 3.2 19.6 44.1 0,274 1.079 18.4 37.2 12.0 49.2 25.9 45.9 1.079 30.7 12.0 0.535 36.3 1.079 16.4 39.8 12.0 0.897 19.9 1.301 13.4 28.1 25.1 20.0 0.274 13.7 34.2 33.9 0,535 22 .o 1.301 21 . o 20.0 30,8 34 .O 1.301 0,897 20.2 18.8 20.0 f3.8 Error f 2 . 8 0,274 10.1 39.5 5 .O 0.699 9.7 37.3 46.9 0,699 11 .o 48.6 0.535 11 2 5.0 50 4 9.7 50.9 0,897 0.699 5 .O 10.3 20.4 13.9 24.8 1.079 15.3 12.0 0.274 24.4 16.1 22.6 1.079 16.3 12.0 0.535 23.2 16.0 1.079 20.7 14.5 12 .o 0.897 13 8 11.7 0,274 7.8 1.301 20.0 6.8 15.9 10.5 16 .O 1.301 20.0 10.6 0.535 11.9 11.3 14.0 1.301 20.0 12.2 0.897 zk2.4 Error f0.9 9.8 11.5 0.274 0,699 4 2 5 .O 4.5 29 .O 27.3 10.2 0,535 10.4 0.699 5.0 0,699 ... ... ... 0.897 5 .O 18.0 1i.i 13.2 20.5 0.274 1.079 12.0 30.8 29.8 1.079 20.5 0.535 12.0 20.5 17.4 15.9 1.079 11.2 0.897 11.5 12.0 5.1 2.8 0.274 2.8 4.3 1.301 20.0 16.3 15.6 10.6 1.301 0,535 20.0 10.9 4.9 2.1 2.0 1,301 0.897 1.8 20.0 fl.? Error & O , 1 ~~
% CdO
5 6 7 8 9
T, 'C 400 400 400 400 400 400 400 400 400
10 11 12 13 14 15 16 17 18
450 450 450 450 450 450 450 450 450
1.88 3.43 7.88 1.88 3.43 7.88 1.88 3.43 7.88
19 20 21 22 23 24 25 26 27
500 500 500 500 500 500 500 500 500
1.88 3.43 7.88 1.88 3.43 7.88 1.88 3.43 7.88
28 29 30 31 32 33 34 35 36
5 50 5 50 550 550 5 50 550 5 50 550 550
1.88 3.43 7.88 1.88 3.43 7.88 1.88 3.43 7.88
1 2 3 4
~~
1.88 3.43 7.88 1.88 3.43 7.88 1.88 3.43 7.88
Graphic constructions of isoyield contours at each temperature were made by first effecting canonical transformation of each of the second-degree equations and then plotting particular solutions obtained for various substituted values of a, the weight % naphthalene yield. Isoyield contours are traced on Figures 11, 12, 13, and 14 to facilitate comparison, interpolation, extrapolation, and generalization. The yield data a t 400" and 550" C. are correlated by equations for ellipses; and a t 450' and 500' C . . by hyperbolas. These changes in the natures of the polynomials indicate pronounced alterations in the reaction mechanisms (9). Complicated variable interactions are apparent. The oxygen content of the catalyst, and consequently its activity and selectivity for the various steps in the multistep reaction sequence, is dependent on the cadmia concentration as well as on the oxygento-feed ratio in the gas phase introduced. Kinetic and mechanistic interpretations are further complicated by the fact that these microreactor conditions are not equilibrium conditions. The isoyield contours for 400' C. are ellipses with predicted yields of up to 80 i 5 weight yo naphthalene, approaching the theoretical limit, indicated at moderate oxygen concentration and nearly 100% CdO. That is, these high yields are obtained at relatively moderate air-to-feed ratios and at long equivalent contact times. But at each cadmia concentration there is an optimum oxygen concentration for the highest yield. Further30
I&EC PROCESS
DESIGN A N D DEVELOPMENT
more, as the cadmia concentration increases, the o q g e n requirement for maximum yield becomes less. O n the other hand, the isoyield contours for 450" C. are hyperbolas, with a region of high yields predicted at slightly lower oxygen levels than a t 400' C. and a t much lower cadmia levels. There is no limitation to approaching theoretical yields. Oxygen levels become yield- or rate-controlling a t very low levels of cadmia; cadmia levels become yield- o r ratecontrolling a t higher levels of cadmia. Since simultaneous three-body reactions are rare, this implies that a t least two consecutive two-body reactions are involved and that they are of similar importance. The fact that yields at high levels of cadmia are oxygen-independent a t lower oxygen concentrations suggests that the rate-controlling step in this domain may be the reaction of 1-methylnaphthalene with oxygen-activated cadmia sites : 1 MN
+C
kl
d O . 0 2 L [ l Mr\;*CdO.Os]$
(14)
The isoyield contours for 500" C. are also hyperbolas, with the indicated high yield region shifted somewhat toward higher cadmia concentrations and lower oxygen concentrations. Here, at fixed cadmia concentrations, the yirld is inversely dependent on oxygen concentrations and, a t fixed oxygen concentrations, is fairly independent of cadmia concentrations.
The concentration of preadsorbed oxygen on cadmia is probably reduced a t higher temperatures. Reaction of the oxygen with the adsorbed 1-methylnaphthalene may be rate-determining in this domain. The yield data a t 550' C. are again fitted by ellipses, suggesting again a marked change of reaction mechanism. Noncatalytic gas-phase oxidation or possibly some thermal cracking in the gas phase may be less selective competing reactions. The converging isoyield contours indicate a maximum yield of 34 I-t 2 weight %;> about one third of the theoretical !ield. This region of operation is impractical for good naphthalene yields. The conversion data a t each temperature level were fitred to a polynomial of the general form p
.4' $. B'x
=
+ C'y + D'x2 + E ' y 2 + F'xy
IO
I3
n 0
a n.
W
z
W
_1
a r Ir L a z
I 3
(15)
naphthalene conversion based on the \\.here P is the weight total 1-methylnaphthalene feed, x is the log (wt. % CdO), a n d y is the log (wt. % 0 2 ) . Data points (Table V I I ) and isoconversion contours are also plotted on Figures 11. 12. 13>and 14. The variance, or lack of fit, is considerably Iclver for all the isoconversion data and decreases with incrrasing temperature. All of the isoconversion data are fitted by ellipses over the entire temperature range 400' to 550' C. The centers of all these ellipses are at approximately the same combination of variables, and the maximum conversion under these conditions is a little over 25 \\.eight yo a t 400' C. and falls off slightly with increasing temperature to a a t 550' C. little over 20 weight With the isothermal yield and conversion contours plotted on the same graphs. it is possible to see the better combinations of reaction conditions. The regions of maximum yields and con\,ersions are summarized in Table T.111. Yields and conversions and the best combinations of these decrease with increasing temperature. The best operating conditions lie in the temperarure region 400' to 450" C .
30
20
loLxtkiA 0
4.0
30
20
ELECTRONEGATIVITY
OF ANION ELEMENT
Figure 10. Comparison of electronegativity values and catalytic activity a t 550" C. At about the 20% combustion level for a11 catalysts
Conclusions 0
Lower homologs are isolatable products in the vapor-phase oxidation of alkylaromatics over many catalysts. Saphthalene is a n isolatable product in the oxidation of alkylnaphthalenes, and benzene is a n isolatable product in the oxidation of alktlbenzenes. There are many catalysts for the oxidation of alkylaromatic hydrocarbons. but they differ markedly in activitv and selectivity. The most active and selective catalyst discovered for naphthalene production is cadmia on silica. Its selectivity depends upon the crystalline nature of the cadmium and associated atoms. activation conditions, reaction temperature, extent of combustion, cadmia concentration, nature of the support. and oxvgen-to-feed ratio. Catalysts and conditions can be adjusted to obtain good yields of these lo\\ er homolog hydrocarbons under microreactor
Table VIII.
I"
O
I0
LOG,,
15
[ W T % CdO]
Figure 1 1. lsoconversion and isoyield contours for selective oxidation of 1 -methylnaphthalene to naphthalene over cadmia a t 400" C.
Summary of Ranges of Optimum Conditions and Results Results -~
wt.%
T, C. 400 450 500 550
0 5
';
naphthalene yield
1.7-7.6 2.4-9.5 2.8-5.8
25 4z 3 20 3~ 2 15 2= 2 20 2
Conditions LVt. CdO W't. 0 2 10.0-22.0 3.4-4.9
6 6-23.5 7.4-13.8 8.0-12.6
+
H't. yo conoerston of
7 .\f-Vto .\' 52-61 i 5 33-58 zk 4 19-36 i 2 31-36 i 2
0
os
I O
LOG,,
15
[WT. % C d O ]
Figure 12. lsoconversion and isoyield contours for selective oxidation of 1 -methylnaphthalene to naphthalene over cadmia a t 450" C. VOL. 3
NO. 1
JANUARY 1964
31
ISOCONVERSION
----, ,
Pi---- /
/---
05
0
I O
LOG,,,
15
05
[WT % CdO]
I O
LOG,,
15
[WT. % C d O ]
Figure 13. lsoconversion and isoyield contours for selective oxidation of 1 -methylnaphthalene to naphthalene over cadmia at 500" C.
Figure 14. lsoconversion and isoyield contours for selective oxidation of 1 -methylnaphthalene to naphthalene over cadmia a t 550" C.
conditions. Preliminary results indicate that optimum conditions for the oxidative demethylation of toluene would not be the same as those found for the oxidative demethylation of 1-methylnaphthalene. However, conditions for the demethylation of toluene may be found by using the gas-liquid chromatograph-microreactor technique, with the aid of experimental design and statistical analysis. Optimum conditions will differ somewhat with different reactors. The microreactor conditions are nonequilibrium conditions. Results would probably be different under equilibrium conditions. Results would probably be different under equilibrium conditions in a large-scale reactor, but there should be qualitative agreement. This microreactor technique, combined with experimental design and statistical analysis, may be used to investigate rapidly either conventional or new reactions for their optimum conditions at tremendous saving in time, laboratory equipment, investment, and scientific manpower. These research procedures should open profitable areas of scientific investigation which heretofore would have been prohibitively expensive for many research organizations.
spectrographic analyses, and Robert F. Sieck for assisting in the experimental research. They also thank Oscar Kempthorne for his advice on the experimental design and statistical analysis techniques employed and H. R. Bailey for programming the polynomial calculations on the digital computer.
Acknowledgment
The authors thank Pauline Dunton and Doris Hulse for x-ray and !vet chemical analyses. .4.L. Schalge for emission
32
l&EC
PROCESS D E S I G N A N D D E V E L O P M E N T
literature Cited (1) Davies, 0. L., ed., "Design and Analysis of Industrial Experi-
ments," pp. 495-578, Hafner, New York, 1956. (2) Doumani. T. F.. Ind. Ene. Chem. 50. 1677-80 (1958). (3j Franklin,". L:, PinchYbeck, P. H., Popper: F.,'Trans. Inst. Chem. Engrs. (London) 34,280-93 (1956). (4) . , Kokes, R. J., Tobin, H., Jr., Emmett, P. H., J . Am. Chem. SOC. 77, 5860-2 (1955). (5) Marisic. M. M.. Zbid.. 62. 2312-7 (1940). (6j Norton,'C. J., Bull. Chem: SOC.Jap& 34,'1545-7 (1961). (7) Norton, C. J., Moss, T. E., IND.ENG.CHEM.PROCESS DESIGK DEVELOP. 2,140-7 (1963). (8) Pauling, L., "Kature of the Chemical Bond," 3rd ed., pp. 88-97, Cornell University Press. Ithaca, N. Y., 1960. (9) Pinchbeck, P. H., Chem. Eng. Sci. 6 , 105-11 (1957). (10) Stein, K. C., Feenan, J. J.. Thompson, G. P., Shultz, J. F., Hofer, L. J. E., Anderson, R. B.: Ind. Eng. Chem. 52, 671-4 (1960). RECEIVED for review January 7, 1963 .ACCEPTED May 31, 1963 Division of Petroleum Chemistry, 144th Meeting, ACS, LOS Angeles, Calif., March 1963.