Chemical Engineering Journal 278 (2015) 183–189

Contents lists available at ScienceDirect

Chemical Engineering Journal

journal homepage: www.elsevier.com/locate/cej

Molecular size distribution in synthesis of polyoxymethylene dimethyl

ethers and process optimization using response surface methodology
Yanyan Zheng, Qiang Tang, Tiefeng Wang ⇑, Jinfu Wang ⇑
Beijing Key Laboratory of Green Reaction Engineering and Technology, Department of Chemical Engineering, Tsinghua University, Beijing 100084, China

h i g h l i g h t s

 Theoretical analysis of PODEn molecular size distribution (MSD) was performed.

 The MSD model was based on a sequential reaction mechanism.
 The MSD model follows the Schulz–Flory distribution.
 Response surface methodology was used to optimize the MSD.

a r t i c l e i n f o a b s t r a c t

Article history: Polyoxymethylene dimethyl ethers (PODEn, CH3O(CH2O)nCH3, where n P 1) are ideal diesel fuel
Available online 24 October 2014 additives. Among the PODEn compounds, PODE3–5 have the best properties as diesel additives. A theoret-
ical analysis of the molecular size distribution of PODEn synthesized from dimethoxymethane (DMM) and
Keywords: paraformaldehyde (PF) was performed based on a sequential reaction mechanism. The molecular size
Polyoxymethylene dimethyl ethers distribution model follows the Schulz–Flory distribution, and showed a good prediction ability at different
Sequential reaction mechanism reaction temperatures (T) and DMM/CH2O mole ratios (M), which verified the sequential reaction
Schulz–Flory distribution
mechanism during the formation of PODEn. The product distribution was optimized using the molecular
Response surface methodology
size distribution model and response surface methodology (RSM). At optimum operating conditions of
T = 105 °C and M = 1.1, the conversion of formaldehyde X CH2 O has a high value of 92.4%, and the fraction
of PODE3–5 in the PODEn mixture is 33.2 wt%, while the fractions of PODEn>5 and PODE2 are 9.4 wt% and
24.3 wt%, respectively.
Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction of methylal. Because water reacts with formaldehyde and PODEn,

the side reactions are notable in this reaction system and make
Polyoxymethylene dimethyl ethers (PODEn, CH3O(CH2O)nCH3, the product purification process very complex [5]. To avoid this
where n P 1) are receiving much attention as ideal diesel fuel problem, Burger et al. [5,6] investigated the synthesis of PODEn
additives, which could significantly reduce the smoke and engine from dimethoxymethane (DMM, namely PODE1) and trioxymeth-
exhaust emissions during combustion [1,2]. The production of ylene, and obtained 23.6 wt% yield of PODE3–4. However, trioxy-
PODEn from C1 chemicals, like methanol and formaldehyde, can methylene is more expensive as a source of formaldehyde and
fully utilize the large surplus of C1 chemicals [3], alleviate the more prone to form PODEn>5 compounds, compared with
diesel supply crisis, and thus bring enormous economic and paraformaldehyde (PF) [8]. In our previous work, we reported the
environmental benefits. synthesis of PODEn from dimethoxymethane (DMM) and PF over
The studies on the synthesis of PODEn are very limited in the cation exchange resins [8]. Under optimized conditions, the con-
literature, and most of them are experimental studies on the version of formaldehyde was 85.1%, and the fraction of PODE3–5
synthesis reaction [4–8]. Zhao et al. [4] studied the synthesis of in the product was 36.6 wt%.
PODEn from methanol and trioxymethylene using molecular sieves In the PODEn compounds, only the PODE3–5 compounds are
as catalysts. In this system, water is produced during the formation ideal diesel additives, because PODE2 does not fulfill the security
criterion due to its low flash point, and PODEn>5 precipitate at
⇑ Corresponding authors. Tel.: +86 10 62794132; fax: +86 10 62772051. low temperatures due to their high melting points [2,6]. Therefore,
E-mail addresses: wangtf@tsinghua.edu.cn (T. Wang), wangjfu@tsinghua.edu.cn the PODE2 and PODEn>5 compounds need to be separated and
(J. Wang).

http://dx.doi.org/10.1016/j.cej.2014.10.056
1385-8947/Ó 2014 Elsevier B.V. All rights reserved.
184 Y. Zheng et al. / Chemical Engineering Journal 278 (2015) 183–189

Nomenclature

NCH2 O;e equilibrium amount of formaldehyde (mol) NCH2 O;R the amount of reacted formaldehyde
kp reaction rate constants of the forward reactions (s1) M DMM/CH2O molar ratio
kd reaction rate constants of the reverse reactions (s1) X CH2 O equilibrium conversion of formaldehyde
N1,e equilibrium amount of DMM (mol) ae dimensionless factor
Nn,e equilibrium amount of PODEn (mol) n polymerization degree of PODEn
N1,0 initial amount of DMM (mol) wn mass fraction of PODEn

recycled to the reactor in an industrial process [6]. The overmuch cation exchange resins were dry resins of H+ type provided by
production of PODE2 or PODEn>5 would significantly increase the Tianjin Bohong Resin Technology Co., Ltd.
energy cost. When recycled into the reactor, a portion of PODEn>5
reacts with formaldehyde and forms larger PODEn compounds, 2.2. Synthesis of PODEn
leading to a complicated separation and recycling process.
Therefore, it is of great importance to study the molecular size dis- The reaction experiments were carried out in a 0.5-L stirred
tribution of PODEn and optimize the process conditions to improve autoclave operated in batch. For each experiment, the mixture of
the molecular size distribution. PF and DMM was first loaded in the reactor. To obtain the equilib-
The molecular size distribution is most studied in polymer rium concentration data, the reaction proceeded until the concen-
chemistry. Knowledge of molecular size distribution is significant trations became unchanged with time on stream.
not only for inferring the properties of polymers [9], but also for The composition of the PODEn mixture was quantitatively
a better understanding of the reaction mechanism [10]. The analyzed by gas chromatography–mass spectrometry (GC–MS).
Schulz–Flory (SF) distribution [11–13] and Poisson distribution The product sample (0.5 mL) was diluted with 5 mL of undecane.
[14–16] are the most widely used theoretical molecular size distri- 1.0 lL of the solution was injected into a Shimadzu 2010 plus GC
butions. Generally, the SF distribution is based on the precondition equipped with an MXT-5 column (5% diphenyl/95% dimethyl poly-
of constant probability of chain growth [17]. In addition to its wide siloxane, 30 m  0.25 mm  0.1 lm) and a flame ionization detec-
applications in conventional processes like polycondensation and tor (FID). The column temperature program comprised two stages:
free radical addition polymerization processes, the SF distribution the initial temperature was set at 40 °C for 5 min, which was then
is also used in Fischer–Tropsch synthesis [18–20] and carbon ramped to 320 °C at 20 °C/min and held for 10 min. Nitrogen was
nano-tube (CNT) preparation [21]. The other theoretical molecular used as carrier gas. An Agilent G2579A MS was used to identify
size distribution, namely Poisson distribution, is considered to be a the species with different residence times in the GC column.
fundamental distribution for the conventional living polymeriza- Because no gas or solid byproducts were formed, the carbon
tion [10,22]. The living polymerization refers to a polymerization balance was good, with error within ±5%. The overall selectivity
of organic molecules when termination is avoided but the true to PODEn>1 was over 98% at the experimental conditions in this
equilibrium is not reached [23]. Zhao et al. [24] found that the work. In the following theoretical analysis, the equilibrium
product distribution follows the SF distribution during the synthe- conversion of formaldehyde (X CH2 O ) refers to the mass fraction of
sis of PODEn, but did not derive the molecular size distribution formaldehyde converted when the system reached equilibrium,
model based on reaction mechanism. and was calculated by:
The response surface methodology (RSM) is a collection of
statistical and mathematic techniques, and has been widely used NCH2 O;feed  NCH2 O;product
X CH2 O ¼ ð1Þ
to develop, improve and optimize a process [25–30]. The use of NCH2 O;feed
RSM highly enhances the experiment and optimization efficiency.
where the amount of formaldehyde (mol) in the feedstock
Based on the designed experimental data, the RSM gave a func-
(N CH2 O;feed ) and in the equilibrium system (NCH2 O;product ) were
tional relationship between the response and the independent
determined by the method provide by ASTM D2194-02 (2012).
variables.
To evaluate the product distribution, a product distribution
In this work, a theoretical analysis of the molecular size distri-
index (PDI) is defined as follows:
bution of PODEn synthesized from DMM and PF was performed
based on a sequential reaction mechanism. During the process PDI ¼ w3—5  h1 wn>5  h2 w2 ð2Þ
optimization in synthesis of PODEn, both the molecualr size distri-
bution and conversion of formaldehyde should be considered. The where h1 and h2 are the relative coefficients of PODEn>5 and PODE2,
optimization of the molecualr size distribution and conversion of respectively. A larger PDI means a compromise result of a higher
formaldehyde was carried out using RSM. To the best of our knowl- yield of target product PODE3–5 and lower yields of the by-products
edge, this is the first report on theoretical analysis of the molecular PODEn>5 and PODE2. Considering that PODEn>5 is more undesirable
size distribution of PODEn and optimization of the molecular size than PODE2 as discussed in Section 1, h1 is set to be larger than h2.
distribution by RSM. The values h1 and h2 can be adjusted according to the process
demand. Herein, h1 and h2 were set as 0.9 and 0.3, respectively.

2.3. Process optimization using RSM

2. Materials and methods
To optimize the process of PODEn synthesis, both the molecular
2.1. Materials size distribution (reflected by PDI) and the conversion X CH2 O should
be considered. In this reaction system, PDI and X CH2 O are deter-
DMM (analytic reagent grade, AR) was purchased from Alfa mined by the reaction temperature (T) and DMM/CH2O molar ratio
Aesar-Johnson Matthey. PF (analytic reagent grade, AR) was (M). While the effect of the reaction time on PDI and X CH2 O will be
purchased from Sinopharm Chemical Reagent Co., Ltd. The NKC-9 investigated in future work on kinetics.
 Y. Zheng et al. / Chemical Engineering Journal 278 (2015) 183–189 185

In the present work, RSM with 5-level and 2-factor central com- 3.2. Equilibrium molecular size distribution model
posite design (CCD) was used to optimize the PDI and X CH2 O . For
two independent variables, the axial parameter is 1.414 for rotat- When the reaction system reached equilibrium, the amounts of
able CCD. The experiment was designed for 5 levels of varying all the species are constant [31], i.e.
independent variables which were coded as 1.414, 1, 0, +1, X
1 X
1
and +1.414, as shown in Table 1. Considering the temperature dNCH2 O;e =dt ¼ kp NCH2 O;e Nn;e þ kd Nn;e ¼ 0 ð7Þ
tolerance of NKC-9 resin, the center value and step size of T are n¼1 n¼2
set as 75 °C and 15 °C, respectively. The center value and step size
of the variable M are 2 and 1, respectively. Multiple regression dN1;e =dt ¼ kp NCH2 O;e N1;e þ kd N2;e ¼ 0 ð8Þ
analysis was used to derive the second-order polynomial equations
to predict the PDI and X CH2 O . A general second-order polynomial dNn;e =dt ¼ kp NCH2 O;e ðNn1;e  Nn;e Þ þ kd ðNnþ1;e  Nn;e Þ ¼ 0
equations [26] can be expressed as: ðn ¼ 2; 3; . . .Þ ð9Þ

PDI ¼ b0;P þ b1;P x1 þ b2;P x2 þ b11;P x21 þ b22;P x22 þ b12;P x1 x2 ð3Þ where N CH2 O;e , N1,e and Nn,e are the equilibrium moles of formalde-
hyde, DMM and PODEn in the liquid phase.
Based on the reaction mechanism, the total amount of PODEn
X CH2 O ¼ b0;X þ b1;X x1 þ b2;X x2 þ b11;X x21 þ b22;X x22 þ b12;X x1 x2 ð4Þ (n P 1) is equal to the initial amount of DMM, therefore
X
1
where x1 and x2 are the coded independent variables, namely the Nn;e ¼ N1;0 ð10Þ
temperature (T) and DMM/CH2O molar ratio (M), respectively; b0;P n¼1

and b0;X are the intercept coefficients; b1;P ; b2;P ; b1;X and b2;X are the
linear coefficients; b11;P ; b22;P ; b11;X and b22;X are the quadratic X
1
Nn;e ¼ N1;0  N1;e ð11Þ
coefficients; and b12;P and b12;X are the interaction coefficients. n¼2

where N1,0 is the initial amount of DMM.

3. Molecular size distribution model Substitution of Eqs. (10) and (11) into Eq. (7) gives
NCH2 O;e ¼ kd ð1  N1;e =N1;0 Þ=kp ð12Þ
3.1. Mechanism for PODEn formation from PF and DMM
Combination of Eqs. (8), (9) and (12) yields
In our previous work, a sequential reaction mechanism was
Nn;e ¼ N1;e ð1  N1;e =N1;0 Þn1 ð13Þ
proposed for the synthesis of PODEn from DMM and PF, as shown
in Fig. 1 [8]. According to this mechanism, the PF molecule The total amount of the added CH2O groups equals to the
depolymerized to formaldehyde monomers, and PODEn1 and amount of reacted formaldehyde (denoted by N CH2 O;R , exclusive of
formaldehyde react to form PODEn. This coincides well with the the formaldehyde converted to by-products as discussed in Section
experimental phenomena that PODE2 was firstly formed, and then 2.2). This gives:
PODE3, PODE4, . . ., PODE7 appeared one by one during the synthe- X1
N1;0
sis of PODEn. This also excludes another mechanism proposed in ðn  1ÞNn;e ¼ NCH2 O;0 X CH2 O ¼ X CH2 O ð14Þ
literature [6] that the PF molecular breaks into segments (CH2O)n, n¼1
M
and (CH2O)n reacts with DMM to form PODEn+1 in one step, which where N CH2 O;0 is the initial amount of formaldehyde, X CH2 O is the
would lead to a simultaneously formation of PODEn components equilibrium conversion of formaldehyde and M is the initial
with different polymerization degree. DMM/CH2O molar ratio.
In addition, it was found that when the reactants were PODE2 For convenience, a dimensionless factor ae is defined as the ratio
and PF, both PODEn>2 and DMM were produced. This indicated that of N CH2 O;R to N1,0 at equilibrium:
the polymerization reactions to PODEn were reversible. Therefore,
the propagation of the polymerization involves the following X CH2 O
ae ¼ NCH2 O;R =N1;0 ¼ ð15Þ
reactions: M
Thus Eq. (14) can be expressed as
kp
DMM þ CH2 O !
 PODE2 ð5Þ X
1
kd
ðn  1ÞNn;e ¼ N1;0 ae ð16Þ
n¼1

kp
PODEi þ CH2 O !
 PODEiþ1 ði ¼ 2; 3; . . . ; nÞ ð6Þ Combining Eqs. (13) and (16) gives
kd
N 1;0
N1;e ¼ ð17Þ
where kp and kd are the rate constants of the forward (polymeriza- 1 þ ae
tion) and reverse (depolymerization) reactions, respectively. Since   n1
the PODEn compounds are homologous series, kp and kd are Nn;e ae ae
¼ 1 ð18Þ
assumed to be independent of n. N1;0 1 þ ae 1 þ ae
This is the equilibrium molecular size distribution of the PODEn
Table 1 products. It follows the SF distribution, where ae/(1 + ae)
Coding of independent variables T and M. corresponds to the probability of chain growth. The weight fraction
distribution could be calculated by
Independent variables Coded level
 
1.414 1 0 +1 +1.414 M DMM þ M CH2 O ðn  1Þ an1
e
wn ¼ ð19Þ
T: reaction temperature 53.79 60 75 90 96.21 M DMM þ M CH2 O ae ð1 þ ae Þn
M: DMM/CH2O molar ratio 0.59 1 2 3 3.41
where wn is the mass fraction of PODEn compound.
186 Y. Zheng et al. / Chemical Engineering Journal 278 (2015) 183–189

Fig. 1. Sequential reaction mechanism for the formation of PODEn from PF and DMM.

In this system, the depolymerization of PF is the rate-determining

step, while the reversible chain growth reactions are in a pseudo-
equilibrium state at any transient moment. Therefore, the molecular
sized distribution is also applicable for transient product distribu-
tion data, which would be further discussed in future work on
kinetics.

4. Results and discussion

4.1. Prediction ability of the theoretical molecular size distribution

model

The natural logarithm form of Eq. (19) is

 
wn ae 1
ln ¼ n ln þ ln ð20Þ
n þ 1:533 ae þ 1 ae ð2:533 þ ae Þ
Eq. (20) shows that the equilibrium mass fractions of the PODEn
compounds are determined by the factor ae and degree of polymer-
ization n. For a specific ae, the molecular size distribution of the
PODEn compounds can be calculated by Eq. (20). Typical results
of the molecular size distribution with different values of ae are
shown in Fig. 2. When the value of ae varied in the range of 0–
2.5, the mass fraction curves of PODE2 and PODE3–5 show single-
peak patterns. The mass fractions of PODE2 and PODE3–5 reach
their maximum value of 25 wt% and 38 wt% at ae of 0.7 and 1.7,
respectively. In contrast, the mass fraction of DMM monotonically
decreased and that of PODEn>5 monotonically increased with
increasing ae.
The parameter PDI was used to evaluate the product distribu-
tion at different ae. The variation of PDI with ae is also shown in
Fig. 2. The PDI curve has its peak value of 17.5 when ae is 1.1.
Fig. 3. Applicability of the theoretical molecular size distribution model at different
Therefore, it is desirable to optimize the operating conditions to DMM/CH2O molar ratios (M).

make the PDI value close to 17.5, which corresponds to a high yield
of PODE3–5 and reasonable by-production of PODEn>5 and PODE2.
Experiments at different T and M were carried out to check the
model prediction ability, and the results are shown in Figs. 3 and 4.
The plots of ln(wn/(n + 1.533)) with respect to n showed a good lin-
ear relationship, with R2 > 0.99. This confirms that the theoretical
model has a good prediction ability for the molecular size distribu-
tion. The calculated value of ae,cal was determined from the slope
(or intercept) of the line between ln(wn/(n + 1.533)) and n. The cal-
culated values of ae,cal from the slope and from the intercept are
consistent within the deviation allowed (±3%), as shown in the
inserted tables in Figs. 3 and 4. The measured mass fractions of
the PODEn compounds are very close to the calculated values.
Fig. 2. Relationship between the molecular size distribution and PDI in PODEn with The experimental value ae,exp was determined by X CH2 O /M accord-
the factor ae. ing to its definition. The experimental and calculated values of ae
 Y. Zheng et al. / Chemical Engineering Journal 278 (2015) 183–189 187

Table 2
Experimental design matrix and results.

Test No. Independent variables Coded variables PDI () X CH2 O (%)
T (°C) M (mol/mol) x1 x2
1 75 2 0 0 4.76 76.2
2 75 2 0 0 4.80 76.2
3 60 3 1 1 1.33 81.0
4 53.79 2 1.414 0 4.33 72.0
5 75 2 0 0 4.76 76.2
6 90 1 1 1 4.76 75.2
7 60 1 1 1 10.65 56.6
8 75 2 0 0 4.80 76.0
9 90 3 1 1 1.10 79.2
10 75 2 0 0 4.72 76.4
11 96.21 2 1.414 0 7.05 90.2
12 75 3.41 0 1.414 0.14 77.5
13 75 0.59 0 1.414 15.59 45.5

Table 3
Coefficient values of response surface model (the correct use of significant figure).

Coefficient Model for PDI (Eq. (3)) Model for X CH2 O (Eq. (4))

Value p-value Value p-value

b0 4.75 <0.0001 76.2 <0.0001
b1 1.01 <0.0001 5.3 0.0009
b2 5.70 <0.0001 9.2 <0.0001
b11 0.55 0.0005 2.9 0.0265
b22 1.57 <0.0001 6.3 0.0003
b12 1.17 <0.0001 5.1 0.0070
R2 0.9987 0.9657
Adjusted R2 0.9977 0.9412
Predicted R2 0.9905 0.7563
Adeq precision 100.131 21.830

Fig. 4. Applicability of the theoretical molecular size distribution model at different

reaction temperatures (T).
Table 4
Analysis of variance (ANOVA) for the response surface models.

Source Sum of Degree of Mean F-value

squares freedom square
Model for PDI (Eq. (3))
Regression 291.14 5 58.23 1038.36
Residual 0.39 7 0.0056
Lack-of-fit 0.39 3 0.13
Pure error 3.121E003 4 7.803E004
Total 291.53 12
Model for X CH2 O (Eq. (4))
Regression 1442.24 5 288.45 39.40
Residual 51.25 7 7.32
Lack-of-fit 51.17 3 17.06
Pure error 0.080 4 0.020
Total 1493.49 12

Fig. 5. Comparison between calculated value ae,cal from Eq. (20) and experimental (3) and (4) are listed in Table 3. For the two models, the predicted
value ae,exp from Eq. (15). R2 is in a reasonable agreement with the adjusted R2, indicating
that the models well described the relationship between the inde-
pendent variables and the response. The p-value was an important
agreed very well, as shown in Fig. 5. These results further validate factor to evaluate the statistical significance of each regression
the sequential reaction mechanism [8] and the molecular size term. The p-values less than 0.05 indicate that the regression terms
distribution mode proposed in this work. are significant [26–27]. All the regression terms are significant,
among which x2 (coded DMM/CH2O molar ratio) is the most signif-
4.2. Experimental results and statistical analysis using RSM icant term due to its least p-value. This indicates that the DMM/
CH2O molar ratio has the most significant effect on X CH2 O . The
For process optimization using RSM, the parameters PDI and statistical significance of the models was tested by the analysis
X CH2 O are the important factors. The experimental design matrix of variance (ANOVA), as shown in Table 4. The sum of squares is
and the experimental results are shown in Table 2. The analysis an index used to estimate the square of deviation. The mean
of the experimental results in Table 2 was performed using the squares are estimated by dividing the sum of squares by degrees
software Design-Expert V8.0.6.1, and the manual regression model of freedom. The model F-value was used to evaluate the overall sig-
was selected to fit Eqs. (3) and (4). The resulted coefficients of Eqs. nificance of the model, in which the calculated value of F should be
188 Y. Zheng et al. / Chemical Engineering Journal 278 (2015) 183–189

due to the low operating temperature, and the higher M means a

lower conversion of DMM. Therefore, only the latter subtriangular
region is feasible and optimal.
To optimize the synthesis of PODEn, both PDI and X CH2 O should
be considered. In the contour of PDI = 17.5, the point (T = 105 °C,
M = 1.1) is optimum, giving the highest X CH2 O of 92.4%.

5. Conclusions

Among the PODEn compounds, PODE3–5 compounds are the

most ideal diesel additives. A theoretical molecular size distribu-
tion model of PODEn from DMM and PF was proposed based on
the sequential reaction mechanism. The product distribution was
optimized using the molecular size distribution model and
response surface methodology (RSM). The following conclusions
can be drawn from the theoretical analysis and experimental
Fig. 6. Contour plots of factor PDI (production distribution index) as a function of results:
reaction temperature (T) and DMM/CH2O molar ratio (M).
(1) The molecular size distribution model followed the SF
distribution.
(2) The molecualr size distribution model showed a good
prediction ability at different T and M, which verified the
sequential reaction mechanism of PODEn synthesis from
DMM and PF.
(3) The molecular size distribution and formaldehyde conver-
sion were optimized using RSM, and the optimum operating
conditions were determined as T = 105 °C and M = 1.1, where
the formaldehyde conversion reached 92.4%, and the frac-
tion of PODE3–5 in the PODEn mixture was 33.2 wt%, while
the fractions of PODEn>5 and PODE2 were 9.4 and 24.3 wt%,
respectively.

References

[1] M. Marchionna, R. Patrini, D. Sanfilippo, A. Paggini, F. Giavazzi, L. Pellegrini,

From natural gas to oxygenates for cleaner diesel fuels, Stud. Surf. Sci. Catal.
Fig. 7. Contour plots of X CH2 O as a function of reaction temperature (T) and DMM/ 136 (2001) 489–494.
CH2O molar ratio (M). [2] B. Lumpp, D. Rothe, C. Pastotter, R. Lammermann, E. Jacob, Oxymethylene
ethers as diesel fuel additives of the future, MTZ Worldwide eMagazine 72
(2011) 34–38.
greater than the F-table value [26–27]. Herein, the model F-values [3] C.J. Yang, R.B. Jackson, China’s growing methanol economy and its implications
for energy and the environment, Energy Policy 41 (2012) 878–884.
of 1038.36 and 39.40 imply that the two models were statistically [4] Q. Zhao, H. Wang, Z. Qin, Z. Wu, J. Wu, W. Fan, et al., Synthesis of
significant. polyoxymethylene dimethyl ethers from methanol and trioxymethylene
The operating region, where T varies from 45 °C to 105 °C and M with molecular sieves as catalysts, J. Fuel. Chem. Technol. 39 (2011) 918–923.
[5] J. Burger, E. Strofer, H. Hasse, Chemical equilibrium and reaction kinetics of the
varies from 0.5 to 4.0, is the feasible optimization range consider-
heterogeneously catalyzed formation of poly (oxymethylene) dimethyl ethers
ing the tolerable temperature of the catalyst and system pressure. from methylal and trioxane, Ind. Eng. Chem. Res. 51 (2012) 12751–12761.
Fig. 6 shows the contour plots of PDI as a function of T and M. It can [6] J. Burger, M. Siegert, E. Strofer, H. Hasse, Poly (oxymethylene) dimethyl ethers
be seen that the increase of T and decrease of M would increase as components of tailored diesel fuel: properties, synthesis and purification
concepts, Fuel 89 (2010) 3315–3319.
PDI. In order to obtain a high yield of PODE3–5 and avoid overmuch [7] M. Arvidson, M.E. Fakley, M.S. Spencer, Lithium halide-assisted formation of
by-production of PODEn>5 and PODE2, the optimum of PDI is polyoxymethylene dimethyl ethers from dimethoxymethane and
around 17.5, as discussed in Section 3.2. When the operating formaldehyde, J. Mol. Catal. 41 (1987) 391–393.
[8] Y. Zheng, Q. Tang, T. Wang, Y. Liao, J. Wang, Synthesis of a green fuel additive
parameters is on the line segment between (T = 78 °C, M = 0.5) over cation resins, Chem. Eng. Technol. 36 (2013) 1951–1956.
and (T = 105 °C, M = 1.1), the PDI reaches the value of 17.5, thus [9] L.J. Fetters, D.J. Lohse, D. Richter, T.A. Witten, A. Zirkel, Connection between
obtaining 33.2 wt% of PODE3–5 compounds in PODEn, while polymer molecular weight, density, chain dimensions, and melt viscoelastic
properties, Macromolecules 27 (1994) 4639–4647.
controlling the mass fraction of PODEn>5 and PODE2 at 9.4 wt% [10] H. Tobita, Molecular weight distribution of living radical polymers, Macromol.
and 24.3 wt%, respectively. Theor. Simul. 15 (2006) 12–22.
The conversion of formaldehyde X CH2 O is another essential [11] J. Wagner, Small-angle scattering from spherical core-shell particles: an
analytical scattering function for particles with Schulz–Flory size distribution,
index. Fig. 7 shows the contour plots of X CH2 O as a function of T J. Appl. Crystallogr. 37 (2004) 750–756.
and M. In general, operation at high T or high M is favorable to [12] W. Radke, G. Litvinenko, A.H.E. Müller, Effect of core-forming molecules on
increase X CH2 O . The synthesis of PODEn from DMM and PF is endo- molecular weight distribution and degree of branching in the synthesis of
hyperbranched polymers, Macromolecules 31 (1998) 239–248.
thermic and reversible, therefore the conversion of formaldehyde
[13] Y.V. Kissin, Molecular weight distributions of linear polymers: detailed
increases with an increase in the reaction temperature and analysis from GPC data, J. Polym. Sci. Pol. Chem. 33 (1995) 227–237.
DMM/PF ratio. In the two subtriangular regions (T = 45 °C, [14] G.J.P. Britovsek, S.A. Cohen, V.C. Gibson, P.J. Maddox, M.V. Meurs, Iron-
M = 2.2)–(T = 55 °C, M = 4.0)–(T = 65 °C, M = 2.8) and (T = 105 °C, catalyzed polyethylene chain growth on zinc: linear a-olefins with a poisson
distribution, Angew. Chem. Int. Ed. 41 (2002) 489–491.
M = 0.5)–(T = 78 °C, M = 2.75)–(T = 105 °C, M = 3.4), the X CH2 O is [15] R. Granek, M.E. Cates, Stress relaxation in living polymers: results from a
over 80%. But in the former region, the reactions are too slow Poisson renewal model, J. Chem. Phys. 96 (1992) 4758–4767.
 Y. Zheng et al. / Chemical Engineering Journal 278 (2015) 183–189 189

[16] W. Lee, H. Lee, J. Cha, T. Chang, Molecular weight distribution of polystyrene enantioselective lipase from Aspergillus niger by solid-state fermentation, J.
made by anionic polymerization, Macromolecules 33 (2000) 5111–5115. Microbiol. 47 (2009) 563–571.
[17] P. Flory, Molecular size distribution in ethylene oxide polymers, J. Am. Chem. [26] K. Somnuk, P. Smithmaitrie, G. Prateepchaikul, Optimization of continuous
Soc. 62 (1940) 1561–1565. acid-catalyzed esterification for free fatty acids reduction in mixed crude palm
[18] A. Tavakoli, M. Sohrabi, A. Kargari, Application of Anderson–Schulz–Flory (ASF) oil using static mixer coupled with high-intensity ultrasonic irradiation, Energ.
equation in the product distribution of slurry phase FT synthesis with Convers. Manage. 68 (2013) 193–199.
nanosized iron catalysts, Chem. Eng. J. 136 (2008) 358–363. [27] M. Jain, V.K. Garg, K. Kadirvelu, Investigation of Cr (VI) adsorption onto
[19] H. Schulz, Short history and present trends of Fischer–Tropsch synthesis, Appl. chemically treated Helianthus annuus: optimization using response surface
Catal. A Gen. 186 (1999) 3–12. methodology, Bioresour. Technol. 102 (2011) 600–605.
[20] J. Patzlaff, Y. Liu, C. Graffmann, J. Gaube, Studies on product distributions of [28] P.D. Patil, V.G. Gude, A. Mannarswamy, P. Cooke, S. Munsonmcgee, N.
iron and cobalt catalyzed Fischer–Tropsch synthesis, Appl. Catal. A Gen. 186 Nirmalakhandan, et al., Optimization of microwave-assisted
(1999) 109–119. transesterification of dry algal biomass using response surface methodology,
[21] R. Zhang, Y. Zhang, Q. Zhang, H. Xie, W. Qian, F. Wei, Growth of half-meter long Bioresour. Technol. 102 (2011) 1399–1405.
carbon nanotubes based on schulz–flory distribution, ACS Nano 7 (2013) [29] K.P. Gopinath, K. Muthukumar, M. Velan, Sonochemical degradation of Congo
6156–6161. red: optimization through response surface methodology, Chem. Eng. J. 157
[22] I.C. Sanchez, Irreversible anionic polymerization kinetics revisited, Ind. Eng. (2010) 427–433.
Chem. Res. 49 (2010) 11890–11895. [30] B. Pereda-Ayo, D. Duraiswami, J.A. Gonzalez-Marcos, J.R. Gonzalez-velasco,
[23] S.C. Greer, Physical chemistry of equilibrium polymerization, J. Phys. Chem. B. Performance of NOx storage-reduction catalyst in the temperature-reductant
102 (1998) 5413–5422. concentration domain by response surface methodology, Chem. Eng. J. 169
[24] Y. Zhao, Z. Xu, H. Chen, Y. Fu, J. Shen, Mechanism of chain propagation for the (2011) 58–67.
synthesis of polyoxymethylene dimethyl ethers, J. Energ. Chem. 22 (2013) [31] W.B. Brown, M. Szwarc, Molecular weight distribution of ‘‘living’’ polymers,
833–836. Trans. Faraday Soc. 54 (1958) 416–419.
[25] F.J. Contesini, V.C.F. da Silva, R.F. Maciel, R.J. de Lima, F.F.C. Barros, P. de
Oliveira Carvalho, Response surface analysis for the production of an