Process Modeling, Simulation, and Technoeconomic Evaluation of

Subscriber access provided by UB + Fachbibliothek Chemie | (FU-Bibliothekssystem)

Full Paper

Process modelling, simulation and technoeconomic evaluation of separation solvents for the Continuous Pharmaceutical Manufacturing (CPM) of diphenhydramine Samir Diab, and Dimitrios I. Gerogiorgis Org. Process Res. Dev., Just Accepted Manuscript • DOI: 10.1021/acs.oprd.6b00386 • Publication Date (Web): 08 Mar 2017 Downloaded from http://pubs.acs.org on March 13, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Organic Process Research & Development is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Organic Process Research & Development

PROCESS MODELLING, SIMULATION AND TECHNOECONOMIC EVALUATION OF SEPARATION SOLVENTS FOR THE CONTINUOUS PHARMACEUTICAL MANUFACTURING (CPM) OF DIPHENHYDRAMINE Samir Diab, Dimitrios I. Gerogiorgis* Institute for Materials and Processes (IMP), School of Engineering, University of Edinburgh, The King’s Buildings, Edinburgh, EH9 3FB, United Kingdom *

Corresponding author: [email protected] (+44 131 6517072)

HIGHLIGHTS • Demonstrated continuous flow synthesis of diphenhydramine employed for kinetic parameter estimation. • Small PFR reactors achieved for continuous flow synthesis of diphenhydramine with/without carrier solvent. • Thermodynamic modelling of solid-liquid (API solubility) and liquid-liquid (LLE) equilibria performed. • High API recovery (81.3%) and low E-factor (31.06) using methylcyclohexane as a separation solvent. • Significant Total Cost savings achievable using methylcyclohexane for continuous separation (−37.3%).

1 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ABSTRACT Continuous Pharmaceutical Manufacturing (CPM) has been shown to provide significant process benefits, including cost savings and increased material efficiencies, over the currently implemented batch paradigm which suffers from poor heat transfer and mixing and high volumes of waste. The continuous flow synthesis of diphenhydramine, a first-generation anti-histamine, has been demonstrated on the laboratoryscale, featuring in-line synthesis and purification in both a carrier solvent and in a neat mixture. This work develops a process model for the continuous upstream processing of diphenhydramine in order to demonstrate the feasibility and viability of this process. The model features the design of plug flow microreactors for the continuous flow synthesis of 100 kg per annum of diphenhydramine (the option of solventless/neat reactions has also been considered explicitly in all stages of our process analysis), heating requirements, and comparison of candidate solvents for API purification via liquid-liquid extraction. Original results indicate small computed reactor volumes and considerable heating requirements for the plant capacity. Chloroform emerges as a potent extraction agent in comparison to other candidate solvents, allowing the highest material efficiency (environmental factor of 3.43, total cost savings of 49.5%), but its high toxicity prohibits CPM use. Methylcyclohexane is the next strongest performer (environmental factor of 31.06, total cost savings of 37.3%) whose chemical properties render it significantly more acceptable for CPM implementation. The present study exhibits the potential for technological innovation available via CPM in an effort to facilitate the transition from batch production methods in the pharmaceutical industry. Keywords: Continuous Pharmaceutical Manufacturing (CPM); Green chemistry; Process design; Separation design; Process simulation; Diphenhydramine

1.

Introduction

Batch processing has been the traditional manufacturing method used by the pharmaceutical industry due to established regulatory protocol for such a mature technology. Advantages offered by batch methods include high precision product quality control, specific batch recall and versatile equipment usage; thus, batch manufacturing methods currently dominate the pharmaceutical industry1. However, batch processing suffers from several drawbacks such as poor heat transfer and mixing efficiencies (potentially leading to unacceptable quality of product and difficulties in scaleup), large inventories of material (incurring a large plant footprint and high capital expenditures), high volumes of waste and intensive labour requirements2. In the past, production costs in the pharmaceutical industry have been considered low enough such that reductions have been unnecessary. However, research and development (R&D) costs for the pharmaceutical industry have been rapidly increasing over previous decades, with the pharmaceutical industry having the highest R&D expenditures of all industrial sectors3 (Figure 1). Capitalised costs of drug product commercialisation have also been historically increasing4. When the duration of clinical trials of potential drugs and product approval is accounted for, pharmaceutical products lose approximately half of their patent life by the time they have reached the market. This leads to significant profit losses for pharmaceutical firms and increasing competition from generics manufacturers5. Manufacturing contributes approximately 30% of overall costs for pharmaceutical enterprises5. Evidently, if savings are made in the manufacturing stages, the performance of pharmaceutical enterprises will be significantly improved. This can be achieved by investigation and development of innovative manufacturing technologies in order to improve process efficiencies. These improvements can allow significant cost savings in order to maintain profitability and sustainability in pharmaceutical firms. Continuous pharmaceutical manufacturing (CPM), a new production paradigm, has the potential to provide such technological innovation. CPM provides several advantages over batch processing, including lower capital and operating costs6, reduced solvent requirements and waste handling7, increased process efficiency8 and reduced plant footprints9. Mixing patterns in flow reactors are much better understood than those in agitated batch vessels; time for development of flow reactors is significantly reduced in comparison to batch vessels. This allows essential drugs to rapidly reach the market for those in need10,11. Implementing continuous flow synthesis allows the application of microscale equipment, particularly microreactors, featuring tubular flow reactors with inner diameters on the millimetre scale. Microreactors benefit from enhanced mixing and mass and heat transfer, safer operation, and reduced footprint12.


Page 2 of 39

Page 3 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Food producers Banks Oil and gas production Healthcare equipment and services General industrials Leisure goods Aerospace and defence Industrial engineering Chemicals Average Electronic and electrical equipment Software and computer services Automobiles and parts Technology hardware and equipment Pharmaceutiucals and biotechnologies

0

20

40

€ (bn)

60

80

100

Figure 1: Global R&D expenditure by industrial sector; data from the EU Industrial R&D Scoreboard (2012).3.

Recent ventures highlight the beginning of the transition from batch to continuous manufacturing of pharmaceutical compounds. In the past two decades, there have been many demonstrations of API production processes which have benefited by changing to a continuous manufacturing paradigm. Table 1 provides a summary of certain APIs whose manufacturing processes have been converted from batch to continuous mode to various extents. Whilst Table 1 does not provide a comprehensive review of demonstrated CPM of APIs (detailed reviews are available in the literature13–15), it is clear that the transition of manufacturing paradigm from batch to continuous processing remains predominantly in the R&D stages (lab scale and pilot plant capacity) with only two CPM processes at production level. CPM is yet to be widely adopted by the risk-averse industry due to the limited technological understanding and demonstration of process feasibility in this emerging field. Additionally, significant financial investments in existing batch-operated plants make such a drastic change in production paradigm an unfavourable prospect16. Process feasibility and viability studies must be conducted before significant time and financial investments for experimental and pilot plant studies are made. However, experimental demonstration is laborious, time-consuming and expensive, with little technological expertise existing on CPM in the engineering community. Recently, there has been significant interest in process systems engineering (PSE)-driven efforts for pharmaceutical process design. The philosophy of PSE is a knowledge-based approach towards effective solutions for process design; the multipurpose, complex nature of pharmaceutical processes renders it suitable for such methods17. However, it is also this complexity that has prevented the widespread adoption of PSE in pharmaceutical process synthesis and design due to the generic methods in which PSE tools have



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

Page 4 of 39

Table 1: API manufacturing processes converted from batch to continuous mode. API

Application

Process Benefits of Converting to Continuous Mode

Level of Development Labscale Demonstration Flow Synthesis

Purification

Formulation + Tableting

Production Level

Process Chemistry

CPM Capacity (kg/y)

Ref.

Ease / Safety CapEx / OpEx of Costs Operation

Aliskiren Hemifurate

Hypertension

✓

✓

✓

✓

✓

✓

—

360.00

Amitriptyline

Antidepressant

✓

✓

X

✓

—

✓

—

4.24

Antibiotic

✓

✓

✓

—

✓

—

✓

Amoxicillin

Company / Institution

Novartis-MIT CCM, USA University of Hanover, Germany

(21) (22)

Unavailable

GSK

(23) (24)

Antimalarial

✓

X

X

—

—

✓

✓

66.64

MPI for Colloids and Interfaces, Germany

AZD6906

Reflux inhibitor

✓

X

X

✓

—

✓

—

Unavailable

AstraZeneca

(25)

Darunavir

HIV drug

✓

✓

✓

—

✓

—

✓

Unavailable

Janssen

(26,27)

✓

✓

X

✓

—

✓

—

19.36

MIT

(28)

✓

✓

✓

✓

✓

✓

✓

18.72

MIT

(29) (30)

Artemisinin

Diphenhydramine

Antihistamine

Fanetizole

Anti-inflammatory

Fluoxetine

Antidepressant

6-Hydroxybuspirone

Psychotropic agent

Ibuprofen

Anti-inflammatory

✓

X

X

✓

—

—

—

79.68

University of Cambridge, UK

✓

X

X

✓

✓

✓

—

11.84

Eli Lilly

(31)

✓

✓

✓

✓

✓

✓

✓

7.36

MIT

(29)

✓

X

X

✓

—

✓

—

Bristol-Myers Squibb

(32)

✓

X

X

✓

—

✓

—

4.32

Florida State University, USA

(33)

✓

✓

X

✓

✓

✓

✓

64.72

MIT

(34) (35)

1,166.64

Leukaemia

✓

X

X

✓

✓

—

—

0.01


Rufinamide

Antiepileptic

✓

✓

X

✓

✓

✓

—

1.76

MIT

(36)

Tamoxifen

Breast cancer

✓

X

X

✓

✓

—

—

432.00


(37)

Telmisartan

Hypertension

✓

X

X

✓

—

—

✓

0.48

Virginia Commonwealth University, USA

(38)

Vitamin D3

Liver failure

✓

X

X

✓

✓

✓

Tokyo Institute of Technology, Japan

(39)

Imatinib

✓

4

ACS Paragon Plus Environment

Unavailable

Page 5 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


been developed for bulk chemicals manufacturing18. A particular obstacle for pharmaceutical process design via PSE is the difficulty in modelling the presence of solids in various process streams19. Nevertheless, a PSE approach to CPM process design has shown significant promise in recent efforts. However, the early development of this field of technology means that data-driven methods are not readily available. Therefore, a knowledge-based approach is more appropriate for the design and development of complex processes20.Theoretical modelling of CPM processes also allows rapid screening of novel CPM processes for the synthesis of candidate active pharmaceutical ingredients (APIs), whilst circumventing experimental costs and labour requirements40. Process models have been implemented to demonstrate the technoeconomic feasibility and viability of CPM for various APIs41–43. Multiparametric analyses of downstream processes44–51 and plantwide designs52–55 for CPM have also been investigated via PSE methodologies. Ultimately, a process should only be implemented in continuous mode if it shows significant benefits over batch operation. Several APIs have recently been established as promising candidates for CPM56. Among these, two promising APIs have already been investigated for their feasibility in a CPM process: ibuprofen, the popular non-steroidal anti-inflammatory drug33,56 and artemisinin, an important antimalarial43,57. A recent plantwide technoeconomic analysis of the CPM of both APIs, with different separation options and production scales showed significant total cost savings and acceptable material efficiencies for pharmaceutical processes58. Another promising candidate API is diphenhydramine, a first-generation antihistamine with hypnotic and antidepressant applications59. The hydrochloride salt of diphenhydramine is the active pharmaceutical ingredient in many popular brand marketed formulations such as Benadryl®, Unisom®, Tylenol®, Zzzquil® amongst others60. The batch production method, which is still in use today, uses dimethylaminoethanol (DMAE), bromodiphenylmethane and solvent61. The continuous flow synthesis of diphenhydramine has been demonstrated via two methods: in reaction solvent and as a neat mixture of ammonium salts without solvent28. The CPM route features inline purification and crystallisation, full-atom economy and flow of molten ammonium salts, fulfilling several of the 12 principles of green chemistry described by the American Chemical Society62. This study clearly demonstrates the benefits of reduction of environmental impact which can be made available via CPM of diphenhydramine. The benefits of using neat mixtures with microreactor technology are well documented in the literature63–66. Recently, a fully integrated continuous flow synthesis with downstream processing of diphenhydramine in a neat mixture has been recently demonstrated as part of a compact, reconfigurable system capable of synthesising other APIs29. The published continuous-flow synthesis route demonstrates the feasibility of CPM for diphenhydramine on the laboratory scale28. In order to realise the benefits of CPM at production level for this important API, further technological and economic feasibility studies must be conducted. The aim of this work is to produce a process model to simulate the upstream processing (synthesis and purification stages) for CPM of diphenhydramine based upon the published continuous flow synthesis. This will involve establishing novel kinetic expressions of the chemical reactions of API synthesis in solvent and in a neat mixture. Relevant physical property estimations are subsequently made where necessary for subsequent unit operation design. Plug flow reactor (PFR) design are next conducted with the calculation of required reactor lengths for appropriate inner diameters and their required heating duties. The design of a continuous API purification stage uses experimental solubility and equilibrium data as well as theoretical methods, with a critical comparison of extraction performances and material efficiencies of different design options. An economic evaluation of the entire process shall then demonstrate the cost savings and improvements in material efficiency which can be realised by implementing a continuous separation train relative to the batch alternative, as well as the improvements realised by neat synthesis relative to the synthesis in carrier solvent. Finally, a critical discussion of the design methods used will examine the technological feasibility of implementation of CPM of diphenhydramine in R&D.

2.

Process Modelling and Simulation

2.1 Flowsheet Development The flowsheet developed in this work is based upon the published continuous flow synthesis of diphenhydramine described by Snead and Jamison28. The CPM route involves the etherification of chlorodiphenylmethane (CDPM) and dimethylaminoethanol (DMAE) to form the API hydrochloride salt. The synthesis reaction is carried out via two methods: (1) in carrier solvent, N-methyl-2-pyrrolidone (NMP), and (2) as a neat mixture. Herein, API syntheses occurring in carrier solvent and in a neat mixture are referred to as “solvent” and “neat” syntheses respectively (Figure 2). 5 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 39

The developed flowsheets for both syntheses are shown in Figure 3. Reagents and carrier solvent are metered by pumps P-101/2/3 into PFR R-101a/b. Aqueous NaOH solution (3 M) is then pumped into the reactor effluent by P-104 to neutralise the API salt, after heating to the reaction temperature in HX-101a/b. The neutralised effluent is then purified to obtain a product containing the API, following heating to the required separation temperature in HX-102.

(a) NMP

+ H (b) No solvent

CDPM

DMAE

API (salt)

Figure 2: CPM route synthesising API salt, diphenhydramine hydrochloride28. (a) Reaction in carrier solvent at T = 180 ⁰C (b) Reaction without carrier solvent at T = 175 ⁰C.

Figure 3: Developed continuous flowsheet for synthesis of diphenhydramine. (a) Process using carrier solvent (b) Process without carrier solvent.

CDPM NMP

a

DMAE NaOH H2O Solvent

F1

P-101

R-101a/b

F3 F2

F4

F7

P-102

F6

F5

F11

HX-102

F12

F14

a

Separation and API recovery

F15

API (organic)

F16

Waste (aqueous)

P-103

P-104

F9

HX-101a/b

F10

F8

F13

2.2 Mass Balance Calculations Accurate design of all unit operations requires a plantwide mass balance. The following assumptions have been made for mass balance calculations of all component streams: • •

The process production scale is 100 kg of API per annum. Chemical reaction only occurs within the reactor, not in any previous or subsequent units or their connecting lines.


Page 7 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


• • •

No reaction occurs between unreacted reagents and subsequent chemicals introduced in the process. Mixtures are homogeneous and ideal solutions; viscosity variations do not affect flow or conversion. Temperature changes throughout the process cause no phase transitions or precipitation of salts.

Pure component densities at elevated reaction temperatures are required for accurate calculation of mixture densities and volumetric throughputs. Coefficients of isothermal expansion (β) have been found in the literature or estimated (where unavailable) in order to calculate component density as a function of temperature according to Eq. (1). Values of β are listed in Table 5. A plantwide mass balance has been conducted. Table 2 and Table 3 and show the mass flowrates of all components in each stream labelled in Figure 3. ρ1 =

ρ0 1 + β(T1 − T0 )

(1)

The viscosity of the reaction mixture for the synthesis without carrier solvent is of concern and indeed affected by API formation; here, we assume that the reaction mixture viscosity variation does not significantly affect PFR flow patterns or reaction conversion. The constant mixture density assumption throughout (and within) the PFR has also been employed; a higher accuracy requires experimental validation of temperature-dependent transport property models and escapes the purpose of this study. Nevertheless, combined theoretical and experimental investigations of organic reaction mixtures under continuous flow (particularly in the case of solvent-free, i.e. “neat” reactions) will greatly enhance our understanding of transport phenomena, and particularly benefit continuous reactor and separator design.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

Page 8 of 39

Table 2: Mass and molar balances for process with carrier solvent. Component (mass basis) F1 F2 F3 F4 CDPM 12.76 0 12.76 0 DMAE 0 0 0 5.61 NMP 0 28.76 28.76 0 API 0 0 0 0 NaOH 0 0 0 0 H2O 0 0 0 0 Chloroform 0 0 0 0 Total 12.76 28.76 41.51 5.61

F5 12.76 5.61 28.76 0 0 0 0 47.13

F6 0.26 0.11 28.76 18.00 0 0 0 47.13

F7 0 0 0 0 25.19 0 0 25.19

Stream (g h-1) F8 F9 0 0 0 0 0 0 0 0 0 25.19 11.66 11.66 0 0 11.66 36.85

F10 0 0 0 0 25.19 11.66 0 36.85

F11 0.26 0.11 28.76 18.00 25.19 11.66 0 83.98

F12 0.26 0.11 28.76 18.00 25.19 11.66 0 83.98

F13 0 0 0 0 0 0 414.90 414.90

F14 0.26 0.11 28.76 18.00 25.19 11.66 414.90 498.88

F15 0.26 0.11 18.24 17.77 0.00 1.64 40.42 78.44

F16 0.00 0.00 10.38 0.23 25.19 10.04 1.61 47.46

Component (molar basis) CDPM DMAE NMP API NaOH H2O Chloroform Total

F5 6.30 6.29 29.01 0 0 0 0 41.60

F6 0.13 0.12 29.01 7.05 0 0 0 36.31

F7 0 0 0 0 62.98 0 0 62.98

Stream (102 mol h-1) F8 F9 0 0 0 0 0 0 0 0 0 62.98 64.71 64.71 0 0 64.71 127.68

F10 0 0 0 0 62.98 64.71 0 127.68

F11 0.13 0.12 29.01 7.05 62.98 64.71 0 163.99

F12 0.13 0.12 29.01 7.05 62.98 64.71 0 163.99

F13 0 0 0 0 0 0 347.57 347.57

F14 0.13 0.12 29.01 7.05 62.98 64.71 347.57 511.57

F15 0.13 0.12 18.40 6.96 0 9.10 33.86 68.57

F16 0 0 10.47 0.09 62.98 55.72 1.35 130.60

Table 3: Mass and molar balances for process without carrier solvent. Component (mass basis) F1 F2 F3 F4 F5 CDPM 13.16 0 13.16 0 13.16 DMAE 0 0 0 5.79 5.79 NMP 0 0 0 0 0 API 0 0 0 0 0 NaOH 0 0 0 0 0 H2O 0 0 0 0 0 n-Hexane 0 0 0 0 0 Total 13.16 0 13.16 5.79 18.95

F6 0.66 0.29 0 18.00 0 0 0 18.95

F7 0 0 0 0 25.99 0 0 25.99

Stream (g h-1) F8 F9 0 0 0 0 0 0 0 0 0 25.99 12.03 12.03 0 0 12.03 38.02

F10 0 0 0 0 25.99 12.03 0 38.02

F11 0.66 0.29 0 18.00 25.99 12.03 0 56.96

F12 0.66 0.29 0 18.00 25.99 12.03 0 56.96

F13 0 0 0 0 0 0 284.85 284.85

F14 0.66 0.29 0 18.00 25.99 12.03 284.85 344.81

F15 0.58 0.26 0.00 9.35 0.00 0.34 250.66 261.19

F16 0.08 0.03 0.00 8.65 25.99 11.69 34.19 80.63

Component (molar basis) CDPM DMAE NMP API NaOH H2O n-Hexane Total

F6 0.33 0.33 0 7.05 0 0 0 7.70

F7 0 0 0 0 64.98 0 0 64.98

Stream (102 mol h-1) F8 F9 0 0 0 0 0 0 0 0 0 64.98 66.76 66.76 0 0 66.76 131.73

F10 0 0 0 0 64.98 66.76 0 131.73

F11 0.33 0.33 0 7.05 64.98 66.76 0 139.43

F12 0.33 0.33 0 7.05 64.98 66.76 0 139.43

F13 0 0 0 0 0 0 330.53 330.53

F14 0.33 0.33 0 7.05 64.98 66.76 330.53 469.96

F15 0.29 0.29 0 3.66 0 1.89 290.86 296.98

F16 0.04 0.03 0 3.39 64.98 64.87 39.67 172.98

F1 6.30 0 0 0 0 0 0 6.30

F1 6.49 0 0 0 0 0 0 6.49

F2 0 0 0 0 0 0 0 29.01

F2 0 0 0 0 0 0 0 0.00

F3 6.30 0 0 0 0 0 0 35.31

F3 6.49 0 0 0 0 0 0 6.49

F4 0 6.29 0 0 0 0 0 6.29

F4 0 6.50 0 0 0 0 0 6.50

F5 6.49 6.50 0 0 0 0 0 12.99

8


Page 9 of 39

2.3 Kinetic Parameter Estimation Kinetic parameter estimation from the published continuous synthesis is required for accurate PFR design. The original publication provides data for the solvent synthesis at a temperature of 180 °C for residence times of 5, 10 and 20 minutes. For the neat synthesis, data is provided at a reaction temperature of 175 °C for residence times of 16 and 32 minutes. In both reactions, an equimolar mixture of CDPM and DMAE are used28. By plotting different functions of reactant concentrations (representing zero-, first- and secondorder reactions) versus residence time, coefficients of determination have been calculated to estimate reaction orders and rate constants for each reaction. The results of the kinetic parameter estimation for both syntheses are shown in Figure 4. A first-order reaction in CDPM has been the most plausible case for the solvent synthesis (R2 = 0.972) compared to zero- (R2 = 0.438) and second-order (R2 = 0.111) reactions. This is due to the steric hindrance of the aromatic groups on CDPM, rendering it less mobile in solution than DMAE; thus only the concentration of CDPM affects the rate of reaction, hence it is first-order in CDPM. It has been estimated that the first-order rate constant for the solvent synthesis has been k1 = 12.36 h-1. For the neat synthesis, a second-order reaction (first-order in CDPM, first-order in DMAE) is most plausible (R2 = 0.896) compared to zero- (R2 = 0.716) and first-order (R2 = 0.778) reactions. In a neat mixture, the mobility of both reagent molecules is restricted due to the absence of carrier solvent; hence the rate of reaction is dependent on the concentrations of both reagents. For the neat synthesis, the secondorder rate constant has been estimated to be k2 = 46.44 L mol-1 h-1 1.0

25

R² = 0.972

0.5

R² = 0.896 20

CCDPM -1 (L mol-1)

0.0

ln CCDPM (−)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


-0.5 -1.0 -1.5 -2.0 -2.5

15 10 5

-3.0 -3.5

0 0.0

0.1

0.2

0.3

0.4

0.0

0.2

0.4

0.6

Time (h) Time (h) Figure 4: Kinetic parameter estimation from experimental data. (a) Synthesis in carrier solvent (NMP) (b) synthesis in neat mixture28.

Due to the limited available experimental data in the demonstrated continuous flow synthesis of diphenhydramine,28 more elaborate methods could not be employed for kinetic parameter estimation. As such, our methods for kinetic parameter estimation were limited to using the coefficient of determination (R2) as a measure of the goodness of fit of each reaction order. A greater number of experimental data points for both synthetic routes (with and without carrier solvent, NMP) would provide higher reliability in reaction order determination, but kinetic parameter estimation is as accurate as possible, given the experimental data available.28

2.4 Plug Flow Reactor Design Plug flow reactors are the most common microreactor type used in R&D for CPM due to their low volume and efficient heat transfer allowed by the small inner diameters. PFRs can also be coiled to reduce equipment footprint despite long required reactor lengths for small inner diameters. It has been assumed that reaction mixtures in the PFRs have been homogeneous and ideal solutions. In the PFRs, radial and axial concentration and temperature gradients have been assumed negligible i.e. perfect mixing and sufficient heating is provided by appropriate reactor design and completely submerging



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 39

the reactor in heat transfer media with adequate circulation21. The required residence time required to produce 100 kg per annum of API has been calculated using the design equation for a PFR: Xf

τi = CCDPM,0 0

dXCDPM −rCDPM

(2)

The estimation of reactor residence times requires values of the approximate attainable conversion of the limiting reagent in each synthesis case, which is CDPM for both syntheses. The published data indicates a maximum conversion of 98% is possible for the solvent reaction, whilst 95% conversion is attainable for the neat reaction28. A summary of the assumptions made for design of PFRs is provided in Table 4. The calculated residence times in each PFR have been then used to calculate the required reactor volume from the required volumetric throughput. Further detail of residence time calculation for both PFRs is provided in Appendix A. Plug flow regime is assumed in all ID cases, as per the evidence given in the experimental precedent:28 it eliminates radial transport gradients and thus increases conversion. A conclusive flow transition quantification is not possible, given the limited data currently available. A transition to laminar flow would adversely impact reaction conversion and heat transfer. Nevertheless, the explicit quantitative validation of this assumption would require a reliable and explicit quantification (rheological modelling) of mixture viscosity, e.g. via an Arrhenius-type model. This in turn implies either mixture viscosity approximation on the basis of pure compound properties, and/or a detailed experimental rheological study of the mixtures. Experimental evidence suggests PFR use is impractical below a minimum internal diameter (2.5 mm). Accordingly, our study calculates the required PFR volumes and corresponding reactor lengths for a spectrum of diameters, in order to illustrate that acceptably small PFR lengths are required for CPM. In the event that plug flow is not satisfied for e.g. the smallest ID, CapEx would not be adversely impacted (the ID would have to be higher, and the length shorter), but OpEx (cooling) would increase. The effect of reactive mixtures thermal expansion has not been considered, as it escapes the scope of this study; to the extent that nonisothermal reactor operation is not encountered, this parameter does not affect economic analyses significantly; in practice, some (expansion/contraction) piping may be required. Temperature and flowrate sensitivity studies require greater resolution (ODE/PDE-level modelling) of reactive transport phenomena, and imply prerequisite temperature-dependent property estimations. Such uncertain theoretical pursuits would increase uncertainty, due to the lack of experimental data.

Table 4: Kinetic parameter estimation results as a basis for reactor design. Synthesis in carrier solvent Synthesis without carrier solvent Reaction Order 1 2 Rate Law −rCDPM = k1 CCDPM −rCDPM = k2 CCDPMCDMAE 180 175 Reaction Temperature (°C) Conversion (%) 98 95 Rate constant ki (units) 12.36 (h-1) 46.44 (L mol-1 h-1) Coefficient of determination (R2) 0.972 0.896

2.5 Material Property Estimation Material properties of all components are required for accurate reactor and heat transfer design and comparison of separation design options. This section discusses estimation methods used where published pure component data has been unavailable. All published and estimated component properties, including candidate solvents for liquid-liquid extraction are provided in Table 5. 2.5.1 Enthalpy of Formation Enthalpies of formation of components involved in solvent and neat syntheses (i.e. CDPM, DMAE and API) are required for accurate calculation of the reactor heating duties. Where values have been unavailable in the literature, an estimation method has been used67:


Page 11 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


∆0 Hf − hf0 = Ni h1i + Mj h2j i

(3)

j

Here, h1i refers to a contribution of primary functional groups on the molecule, whilst h2j refers to secondary functional group contributions, weighted by the number of occurrences of each functional group (Ni and Mj respectively). For these calculations, hf0 has a constant value of 10.835 kJ mol-1 67.

2.5.2 Specific Heat Capacity at Constant Pressure Specific heat capacities at constant pressure for relevant components are also required for accurate heat transfer design. Where published values have been unavailable, the following group contribution method has been used68. Here, ai, bi, ci and di are constants associated with functional group i, whilst ni is the occurrence of each group on the molecule. Cp = ni ai + ni bi T + ni ci T2 + ni di T3 i

i

i

(4)

i

2.5.3 API Solubility Estimation of API solubility in carrier and extraction solvents is required where published data is unavailable for accurate design of API purification. The UNIFAC method has been used69 using estimated activity coefficients70, described in further detail in Appendices C and E. Estimation of activity coefficients also required calculation of the enthalpy and entropy of fusion of the API, which has been done by a group contribution method 71. Details of this method are provided in Appendix B.

Table 5: Component physical properties. Component

MW (g mol-1)

Density (g cm-3)

Melting Point (⁰C)

Boiling Point (⁰C)

β (x 103 K-1)

Cp (J mol-1 K-1)

∆0Hf (kJ mol-1)

1.11172 1.09072 0.838 a 1.11172 n.r. n.r. n.r. n.r. n.r. n.r. n.r. n.r. n.r.

42.3768 26.6668 167.36 a 59.4768 59.4573 75.37 a 194.77 a 226.64 a 156.00 a 184.50 a 102.30 a 125.34 a 172.50 a

88.3767 67 −79.19 n.r. 67 −60.84 n.r. n.r. n.r. n.r. n.r. n.r. n.r. n.r. n.r.

Process Material

CDPM 202.68 1.14 16.0 140.0 DMAE 89.14 0.89 −59.0 134.1 NMP 99.13 1.03 −24.0 203.0 API 255.36 1.20 162.0 343.7 NaOH 40.00 2.13 318.0 1388.0 H2O 18.02 1.00 0.0 100.0 LLE n-C6H14 86.18 0.65 −95.0 68.0 Solvents n-C7H16 100.21 0.68 −90.6 98.4 CyHex 84.16 0.78 6.5 80.7 MeCyHex 98.19 0.77 −126.3 101.0 DCM 84.93 1.33 −96.7 39.6 TCM 119.37 1.49 −63.5 61.2 Et2O 74.12 1.11 −116.3 34.6 a Published data n-C6H14 = n-Hexane; n-C7H16 = n-Heptane; CyHex = Cyclohexane; DCM = Dichloromethane; TCM = Chloroform; Et2O = Diethyl ether.

MeCyHex = Methylcyclohexane;

2.6 Heat Transfer Design Isothermal operation is assumed for all unit operations i.e. constant temperatures are maintained by provision of sufficient heat transfer requirements. Enthalpies of reaction have been calculated using Hess’ Law, considering the sum of enthalpy changes in cooling the reactants from the reactor temperature to the standard temperature, the standard reaction enthalpy and the enthalpy associated with heating the products from standard to the reaction temperature:



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

∆Hrxn = Cp

reagents

(T0 − T) + ∆H0rxn + Cp

products

(T − T0 )

Page 12 of 39

(5)

The standard reaction enthalpy is calculated as the difference between the sums of enthalpies of formation of products and reagents: ∆H0rxn = ∆H0f products − ∆H0f reagents

(6)

2.7 Purification Design The product stream of both PFRs contains API in an aqueous solution after addition of aqueous NaOH. In the solvent synthesis, the PFR outlet is a binary mixture of NMP and water, whilst in the neat synthesis the aqueous solution is water only. In both solvent and neat cases (CPMa and CPMb, respectively), the aqueous mixture exiting R-101 is assumed to contain several dissolved solutes (unreacted CDPM and DMAE, API and NaOH). API must be extracted from this stream in order to obtain a purified product for solid API removal downstream (not considered here). For CPMa, this can be done by continuous liquidliquid extraction (LLE) where a solvent which is immiscible with the aqueous phase is added to form a multiphase ternary mixture. The API should be preferentially soluble in the added solvent so that extraction is favourable. For CPMb, addition of n-hexane to the aqueous stream yields a binary phase mixture which can be separated by a Zefluor membrane (yield of the organic phase by the membrane is assumed to equal 88%), as in the original CPM publication28. The performance of the membrane with other solvents is unknown, and thus other options for CPMb cannot be explored here. For accurate design of LLE, equilibrium data is required for ternary systems containing NMP, water and solvent (CPMa) and for the binary system of n-hexane and water (CPMb). A Pfizer solvent selection guide, classifying solvents as “preferred”, “useable” or “undesirable”74, has been consulted for solvent selection. A wide range of candidate solvents have been considered for the design of the API purification stage (Table 6). In order for a solvent to be considered suitable for application, three criteria had to be fulfilled: • •

•

The API must exhibit sufficient solubility in the candidate solvent; where possible, experimental data has been compared to the solubility calculated using the UNIFAC model. The candidate solvent must form a two-phase mixture with a wide region of immiscibility. Experimental liquid-liquid equilibrium data is used where possible, with theoretical modelling employed where experimental data is unavailable. If there is no experimental data or required interaction parameters for the models used, the solvent could not be considered. The toxicity of the solvent is considered unacceptable if it is both “undesirable” in the Pfizer solvent selection guide 74 and a Class 1 solvent according to FDA guidelines75.

Theoretical modelling of liquid-liquid equilibria of different ternary systems than those available in the literature has been implemented using the popular UNIFAC69,70 and NRTL76 activity coefficient models, depending on which has been able to converge for different systems. These models are described in Appendices C and D. Organic and aqueous phase composition are estimated via the UNIFAC/NRTL model to conduct a mass balance around the LLE process. The partition coefficient of API between the phases is assumed to be equal to the ratio of API solubilities in each phase.


Page 13 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Table 6 shows the initial list of candidate solvents considered for application. n-Hexane is used in the original CPM publication28 so it is considered here. However, n-hexane is listed as an “undesirable” solvent by the Pfizer selection guide; the authors recommend substitution with n-heptane, which is “preferred”, and thus is also considered here. Cyclohexane is listed as a “usable” solvent and the required interaction parameters for both UNIFAC and NRTL models are available in the literature77,78. Methylcyclohexane can also be considered due to its molecular similarity to cyclohexane; thus both cyclohexane and methylcyclohexane are considered. The performance of chloroform has been recently presented for extraction of the API in this process79 and has been also considered for application in the original continuous flow synthesis publication28. The Pfizer solvent selection guide recommends substituting chloroform with dichloromethane74; thus both dichloromethane and chloroform are considered here. Published liquid-liquid equilibria for the ternary systems containing dichloromethane and chloroform are used here, as the lack of required interaction parameters for both UNIFAC and NRTL models makes theoretical modelling unavailable. Experimental data for the ternary systems NMP + water + carbon tetrachloride, isobutanol, 1-pentanol and 1-hexanol are available80,81. Carbon tetrachloride is strictly controlled by regulatory bodies due to its toxicity and inherent environmental hazard potential, and is not considered here82. Isobutanol, 1-pentanol and 1-hexanol have narrow regions of immiscibility and have been not considered suitable for application. Use of other solvents (alcohols, acids, ketones, esters) has been initially considered, but has been discounted for various reasons (Table 6). The candidate solvents selected for evaluation and their source of liquid-liquid equilibrium data are listed in Table 6. An important design parameter in liquid-liquid extraction performance is the solvent-to-feed (S:F) ratio, as it directly affects the extraction performance, the material efficiency and the operating costs associated with material usage and waste disposal. There are a wide variety of green chemistry metrics which can be used for quantitative comparison of processes83. The simplest, yet most intuitive of these metrics, is the environmental-factor (E), defined as the mass of waste generated per kilogram of product obtained84. E =

mwaste mbpd + mur + mus + muAPI = mAPI mAPI

(7)

For the pharmaceutical industry, the E-factor can reach as high as 20085. This is due to the complexity of typical product molecules which require multistep syntheses and intermediate separations, incurring greater volumes of waste than other processes86. Another useful green chemistry metric is the Mass Productivity (MP), which expresses how efficiently a process uses material, including reagents as well as required neutralisation agents and extraction solvents83, described as follows:

MP =

100 E+1

(8)

API extraction, E-factors and mass productivities have been evaluated for varying operating temperatures and S:F = 0.25-5 (mass basis) for the candidate solvents listed in Table 6. Solubility calculations use the UNIFAC method described in Appendix E.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

Page 14 of 39

Table 6: Initial list of candidate solvents for LLE application Class of Solvent

Alkanes

Chloroalkanes

Alcohols

LLE Solvent

Liquid-Liquid Equilibria

Solubility Data

Classification

Liquid-liquid equilibrium data

Consider for application?

UNIFAC

NRTL

Exp. data

UNIFAC

Exp. Data

FDA

Pfizer

n-C6H14

✓

✓

(87)

✓

X

2

Undesirable

NRTL

✓

n-C7H16

✓

X

X

✓

X

3

Preferred

UNIFAC

✓

CyHex

✓

✓

X

✓

X

2

Useable

NRTL

✓

MeCyHex

✓

X

X

✓

X

2

Useable

UNIFAC

✓

DCM

X

X

(87)

✓

X

2

Undesirable

Experimental data

✓

TCM

X

X

(87)

✓

X

2

Undesirable

Experimental data

✓

MeOH

X

X

X

✓

(88)

2

Preferred

Lack of data

X

EtOH

X

X

X

✓

(88)

3

Preferred

Lack of data

X

1-PrOH

X

X

X

✓

X

3

Preferred

Lack of data

X

1-BuOH

X

X

X

✓

X

3

Preferred

Lack of data

X

iBuOH

X

X

(87)

✓

X

-

Preferred

Single phase mixture

X

1-PnOH

X

X

(80)

✓

X

3

Useable


X

1-HxOH

X

X

(81)

✓

X

-

-


X

Carboxylic acids

AcOH

X

X

X

✓

X

3

Useable

Lack of data

X

Ketones

DMK

X

X

X

✓

(88)

3

Preferred

Lack of data

X

MEK

X

X

X

✓

X

3

Preferred

Lack of data

X

EtOAc

X

X

X

✓

(88)

3

Preferred

Lack of data

X

Esters

X X X X X ✓ iPrOAc 3 Preferred Lack of data n-C6H14 = n-Hexane; n-C7H16 = n-Heptane; CyHex = Cyclohexane; MeCyHex = Methylcyclohexane; DCM = Dichloromethane; TCM = Chloroform; MeOH = Methanol; EtOH = Ethanol; 1-PrOH = 1-propanol; 1-BuOH = 1-butanol; iBuOH = Isobutanol; 1-PnOH = 1-pentanol; 1-HxOH = 1-hexanol; AcOH = Acetic Acid; DMK = Acetone; MEK = Methyl ethyl ketone; EtOAc = Ethyl acetate; iPrOAc = Isopropyl acetate

14


Page 15 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


2.8 Economic Analysis An approximate costing of the process is required to demonstrate the economic benefits and viability of CPM. It has been assumed that the design is constructed at an existing pharmaceutical manufacturing site with infrastructure already in place, operating for 8,000 hours per year. In order to demonstrate the cost savings benefits of CPM separations relative to the batch alternative (BX), an economic comparison of the two processes is made. CPMa and CPMb are also compared to each other to demonstrate the savings associated with using neat synthesis. 2.8.1 Capital Expenditure (CapEx) Where possible, vendor prices have been found for equipment of the same or similar capacity. A costcapacity correlation has been used where equipment of different scales or capacities have been found89. S2 n C2 = C1 f S1

(9)

The exponent, n, is particular to specific equipment and ranges between 0.0-1.0. The correction factor, f, accounts for differences in equipment operation compared to the reference equipment. Where the applicable capacity range of the reference equipment has been much greater than required, the correction factor is not required (i.e. f =1)89. The cost of batch equipment of the same capacity as the continuous equipment is accounted for by applying a factor of 0.9 to the cost of the reference continuous equipment. This is due to the established status of most batch-mode technologies. Inflation of equipment prices taken from the past has been accounted for by applying chemical engineering plant cost indices (CEPCI). For the present, the CEPCI has been taken to equal 578.4. The inflation adjusted total equipment costs provides the Free-on-Board (FOB) cost. FOB cost estimates of plant equipment for the process are provided in Table 7. To calculate the Battery-Limits-Installed-Cost (BLIC), the Chilton Method90 has been employed as follows. The cost of equipment installation is equal to 0.43 times the FOB cost. Additional costs incurred by process piping and instrumentation are taken as 0.3 and 0.12 times the installed equipment costs, respectively. The sum of installed, process piping and instrumentation costs gives the total physical plant cost, to which a construction factor of 0.3 is added (accounting for electricity installations, required buildings, site preparation etc.) to give the BLIC. The Working Capital cost is calculated as 35% and 3.5% of annual material costs for batch and continuous processes, respectively41. Contingency costs are taken as 20% of BLIC. The final CapEx value is taken as the sum of BLIC and working capital and contingency (WCC). The equipment requirements (reactors, heating/cooling and auxiliaries) for API synthesis for CPMa/b are the same but with different capacities. A modular plug-flow microreactor system with integrated heating is used for both reactors (R-101a/b)91. An exponent of 1.0 is used, as suggested in the literature for PFRs89. The cost of heaters and coolers (HX-101a/b, HX-102) are based upon data from an appropriate vendor92. Solenoid metering pumps are used for pumping of reagents and solvents; costing data is taken from vendor sources of such pumps with appropriate capacities93. The batch separation train (BX) for the solvent synthesis case is shown in Figure 5. The original CPM synthesis of diphenhydramine described a threefold batchwise extraction of API with diethyl ether (Et2O) for the solvent synthesis28. Thus, three LLE tanks are required (BX-T101/2/3) with intermediate pumps (BX-P101/2/3) for the batchwise extraction. Pumps are costed using the same source as those used in the API synthesis; LLE tanks are costed as closed, agitated vessels89. The CPM separation train for the solvent synthesis (CPMa) is shown in Figure 5. In this work, a continuous LLE process is designed, with various solvents compared for performance in terms of API recovery and mass efficiency (E-factor). The continuous extraction unit (CPMa-T101) is costed as a mixing tank with a gravity phase separator including the required drives and pumps89. The CPM separation train for the neat synthesis (CPMb) is shown in Figure 5. The original CPM synthesis of diphenhydramine describes extraction of API from the aqueous PFR effluent with hexanes followed by membrane separation of the biphasic mixture (CPMb-M101)28. Costing of the Zefluor membrane used data for a similar membrane94.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

Page 16 of 39

Table 7: Free-on-Board (FOB) cost estimates for process equipment: solvent batch (BX), solvent continuous (CPM a) and neat (CPM b) cases. BX Item Type Ref. Cost Ref. Year f Capacity basis Ref. capacity Design capacity n (GBP) (%) R-101a Reactor 103,208 2014 1.06 Volume (mL) 80.00 23.06 1.00 P-101/2/3/4 Pump 958 2015 HX-101a Heater 4,550 2007 HX-102 Cooler 3,454 2015 BX-P101/2/3 Pump 958 2015 BX-T101/2/3 LLE Tank 24,700 2007 10.33 Turbine Power (kW) 5.00 0.7 0.30 CPM a Item R-101a P-101/2/3/4 HX-101a HX-102 CPMa-P101 CPMa-T101 CPM b Item R-101b P-101/3/4 HX-101b HX-102 CPMb-P101 CPMb-M101

Type Reactor Pump Heater Cooler Pump LLE Tank

Type Reactor Pump Heater Cooler Pump Membrane

Ref. Cost (GBP) 103,208 958 4,550 3,454 958 24,700

Ref. Year

Ref. Cost (GBP) 103,208 958 4,550 3,454 958 5,280

Ref. Year

2014 2015 2007 2015 2015 2007

2014 2015 2007 2015 2015 2007

f (%) 1.06 10.33

f (%) 1.06 10.33

CEPCI 576.1 525.4 525.4

Capacity basis

Ref. capacity

Design capacity

n

CEPCI

Volume (mL) Turbine Power (kW)

80.00 5.00

18.45 0.70

1.00 1.00

576.1 525.4 525.4

Capacity basis

Ref. capacity

Design capacity

n

CEPCI

Volume (mL) Area (m2)

80.00 1.00

3.22 0.13

1.00 1.00

576.1 525.4 525.4

16


Present Cost (GBP) 30,189 958 3,422 3,109 862 14,970

No. Units 1 4 1 1 3 3 Total

FOB (GBP) 30,189 3,832 3,422 3,109 2,587 44,910 88,773

Reference

Present Cost (GBP) 24,151 958 3,802 3,454 958 16,633


FOB (GBP) 24,151 3,832 3,802 3,454 958 16,633 52,830

Reference

Present Cost (GBP) 4,214 958 3,802 3,454 958 852


FOB (GBP) 4,214 2,874 3,802 3,454 958 852 11,743

Reference

(91) (93) (92) (92) (93) (89)

(91) (93) (92) (92) (93) (89)

(91) (93) (92) (92) (93) (94)

Page 17 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(a)

API

Et2O

API

Et2O

Et2O

API Aqueous waste

F12

BX-P101

BX-T101

BX-T102

BX-P102

BX-P103

BX-T103

(b)

(c)

Figure 5: Different separation trains considered for economic analysis. (a) Batch separation train (b) CPM separation train for solvent reaction (CPM a) (c) CPM separation train for neat reaction (CPM b).

2.8.2 Operating Expenditures (OpEx) Specific material prices have been sourced to estimate the costs associated with reagents and solvent requirements, which constitutes a significant portion of OpEx costs; material requirements for the process with different separation options are provided in Table 8; material prices are provided in Table 9. Costs of utility requirements are taken as 0.96 GBP kg-1 of material input. Costs of waste disposal are taken as 0.35 GBP L-1 solvent, which are a significant portion of the waste. These costings are taken from an economic analysis of an integrated CPM process41. Labour costs for CPM separations have been not included due to the small scale of this design, as well as the automated nature of CPM processes95. Batch processes tend to be laborious with manual intervention required for equipment switchover, cleaning and maintenance, and so this should be accounted for in future work. Table 8: Material requirements (kg y-1) for the CPM process with different separation options. Material BX CPMa n-C6H14 n-C7H16 CyHex MeCyHex DCM TCM CDPM 128 169 194 116 126 104 103 DMAE 56 74 85 51 55 46 45 NMP 288 380 437 261 284 234 233 NaOH 252 333 383 228 249 205 204 H2O 117 154 177 106 115 95 95 Solvent 1,181 5,479 6,381 3,804 4,144 341 340


CPMb 203 89 0 400 185 439


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 18 of 39

Table 9: Prices of materials used in the CPM process with different separation options Process materials LLE solvents Material Price (GBP kg-1) Material Price (GBP kg-1) CDPM 0.18 n-C6H14 0.49 DMAE 0.45 n-C7H16 0.13 NMP 1.98 CyHex 1.07 NaOH 0.25 MeCyHex 0.12 H2O 0.60 DCM 0.12 TCM 0.11 All prices quoted from Available Chemicals Directory Et2O 1.59

2.8.3 Total Costs The total cost has been estimated as the net present value (NPV), assuming a 20-year plant lifetime and discount rate (taken as the annual discount rate, i) of 5%. This is a rather conservative estimate for the current economic climate. The NPV accounts for investment in equipment and total operating costs in present-day terms. NPV = CapEx + 20

i=1

3.

OpEx (1 + r)i

(10)

Results

3.1 Mass Balances Figure 6 illustrates component mass flowrates of key streams in Table 2 and Table 3. It can be seen for both syntheses that the API formed constitutes a significant portion of the total stream; hence a large amount of neutralising agent (NaOH) is required for both syntheses to ensure the API is not present as a salt. The amount of NaOH solution required has been scaled up appropriately based upon the amounts added for both syntheses in the original publication28. Approximately the same amounts of solution are added to both PFR effluents (shown in F11 in Figure 3) as the same amount of API salt is being neutralised in each case. The difference in extraction solvent added between the solvent and neat syntheses are associated with differences in extraction efficiency of chloroform and n-hexane, as well as the inefficiency of the Zefluor membrane required for CPMb. 3.2 Reactor Design Table 10 shows the computed reactor sizes for both syntheses. The reactor volumes obtained for both syntheses are small for the target capacity, which demonstrates the benefit of reduced equipment sizes offered by CPM. The significant difference in reactor volume between the solvent and neat syntheses is due to reaction mixture for the latter contains no carrier solvent. Reactor internal diameters of 2.5, 5.0, 10.0 and 15.0 mm have been considered to calculate different reactor lengths corresponding to the required volumes. These are relatively small diameters compared to those considered in the literature, chosen to ensure radial temperature and concentration gradients are negligible21. The diameters considered for the neat reaction are comparable with those reported in similar works for syntheses with neat mixtures (2-6 mm)66. The inner diameter of the reactor is an important tuning parameter for the resulting reactor length and fouling and reactor clogging issues where solid handling is required. It is particularly important to maintain a constant reactor temperature in the neat synthesis to ensure reactor clogging does not occur. Such issues require the need for rigorous process control to ensure high reactor performance55. The original paper describing the continuous flow synthesis of diphenhydramine also describes two side reactions: the formation of benzhydrol by nucleophilic substitution of CDPM, and the selfetherification of benzhydrol to form dibenzhydryl ether28. The work did not provide enough information to estimate kinetic parameters of these side reactions, so the model described here only accounts for the synthesis reaction. In order to obtain more accurate reactor sizes, data on these side reactions must be found in order to obtain accurate overall reaction kinetics.


Page 19 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48


Reactor Outlet

Reactor Feed (a) CDPM

F5

12.76

API

0

NaOH

0

H2O

0

TCM

0

0.11 28.76

28.76

18.00

18.00

25.19

25.19

11.66

11.66

0

0

100

200

300

41.99

0

100

(b)

200

5.79

NMP

0

API

0

NaOH

0

H2O

0

nHex

0

0

0

100

200

F14

0.66

0.66

0.29

0.29

0

0 18.00

18.00

25.99

25.99

12.03

12.03

0

100

200

300

0

300

Organic Product Phase

F11

F5 13.16

DMAE

300

Reactor Outlet

Reactor Feed CDPM

F14 0.26

0.11

28.76

NMP

F11

0.26

5.61

DMAE

Organic Product Phase

284.85

100

200

300

0

Figure 6: Component mass flowrates of key streams in the CPM process (g h-1) (a) with carrier solvent (NMP) (b) as a neat mixture.

19


100

200

300


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 20 of 39

Table 10: PFR design results. CPM Route

Temperature (⁰C)

Total Mass Flow (g h-1)

Total Molar Flow Conversion Volume ID L (mol h-1) (%) (mL) (mm) (mm)

Solvent

180

47.13

0.42

98

18.45

Neat

175

18.94

0.13

95

3.22

2.5 5.0 10.0 15.0 2.5 5.0 10.0 15.0

3759 940 235 104 656 164 41 18

3.3 Heat Transfer Design Calculated reaction enthalpies (Table 11) show that the reaction is endothermic, thus heating is required for both PFRs. The heating requirement of each reactor is considered as the sum of feed heating from standard to reaction temperature and heating of the endothermic reaction. The duty for the neat synthesis is significantly lower than for the solvent synthesis due to the lower throughput, however both are of the same order of magnitude. As a result of the significantly lower reactor volume, the specific duties for the neat PFR are higher than for the solvent PFR. The magnitudes of both duties are considerable for the reactor size and plant capacity; this is due to the elevated reaction temperatures of both synthesis routes. In the original CPM publication, there has been not enough data available to determine the temperature dependence of the reaction rate constants, so analysis into the effect of temperature on conversion and heat transfer design could not be evaluated. It is also assumed that these temperatures are sufficient to keep the reaction mixtures fluid enough such that reactor clogging does not occur, as described in the work of Snead and Jamison28. These calculations consider the power requirements of the reactors only, and do not include pumping, losses, LLE solvent and crystallisation heating/cooling requirements, or power for analysis and control. It is important to consider heat integration of the full process for the detailed design of CPM plants. Table 11: Heat transfer requirements of PFRs. CPM

Reaction

∆Hrxn (kJ mol-1)

Trxn (°C)

Flow in (mol h-1)

Solvent

Endothermic

32.02

180

0.42

Neat

Endothermic

32.07

175

0.13

ID (mm) 2.5 5.0 10.0 15.0 2.5 5.0 10.0 15.0

L (mm) 3759 940 235 104 656 164 41 18

Specific Duty (W cm-1) 0.021 0.084 0.338 0.763 0.023 0.093 0.373 0.850

Power (W) 7.934

1.530

3.4 Purification Design Figure 7 shows the various ternary phase diagrams obtained for the NMP-water-solvent system using theoretical methods or experimental data where appropriate (Table 6). The binary phase diagram for the nhexane-water system required for CPMb has been obtained from experimental data96. Temperatures of 20, 30 and 40 °C have been considered for LLE; higher temperatures have been not considered due to limitations of the UNIFAC liquid-liquid equilibria model97. Liquid-liquid equilibria for systems containing dichloromethane and chloroform could only be assessed at 20 °C as experimental data at other temperatures has been unavailable, and these systems could not be modelled with UNIFAC or NRTL models due to the lack of available interaction parameters. For CPMb, the performance could only be examined at 20 °C as the behaviour of the membrane at higher temperatures is unknown. API recoveries, E-factors and Mass Productivities have been calculated for various S:F values (0.25, 0.50, 1, 2, 3, 4, 5 on a mass basis) for all candidate solvents to evaluate performance on the basis of API extraction and material efficiency. Extraction performance results for both CPMa/b are shown in Figure 8 (see also Table and Table in Appendix F).

20 Environment ACS Paragon Plus

Page 21 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 7: Liquid-liquid equilibrium phase diagrams for different candidate LLE solvents. (a) – (f): Ternary systems of NMP-H2OSolvent, (g) Binary system of n-C6H14 and H2O for the neat process.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

From Figure 8 it can be seen that the extraction performance of all candidate solvents decreases as temperature increases. The best extractions for all solvents occurs at T = 20 °C. It can also be seen that as S:F increases, the API extraction performance of all solvents increases, accompanied by an increase in Efactor. The increase in E-factor indicates a decrease in material efficiency of the extraction process as more solvent becomes present in the aqueous waste stream. In CPMa, both n-hexane and n-heptane achieve unacceptably low API extractions of 60.6% and 53.1% respectively, even at T = 20 °C and a S:F = 5. These poor extraction performances are also accompanied by high E-factors of 40.31 and 48.15, respectively. Cyclohexane and methylcyclohexane achieve acceptable extraction performances of 88.3% and 81.1%, respectively, both at S:F = 5, accompanied by adequate E-factors of 27.18 and 31.06, respectively. Dichloromethane and chloroform both achieve very high extraction performances of 98.5 % and 98.7 %, respectively, at a low S:F = 0.5. Expectedly, the resulting E-factors attained are very good, at 3.59 and 3.43 respectively. Despite their associated toxicity, dichloromethane and chloroform are comparably the best candidate solvents in terms of their extraction performance and material efficiency. The overall extraction performance in CPMb, which includes the inefficiency of the membrane, is unacceptably low, even at S:F = 5. This is partly due to the poor extraction performance of n-hexane, as described above. However, the E-factor is still acceptable at this S:F, which could allow for greater amounts of solvent to be added if necessary. Figure 9 compares the E-factors and Mass productivities obtained for the operating conditions considered here. Expectedly, as S:F increases, MP decreases due to increasing amount of waste incurred by additional use of extraction solvent. MP is directly related to the E-factor, and thus the highest MP corresponds to using chloroform as an extraction solvent at T = 20 °C and S:F = 5. The best performances of all candidate solvents for CPMa/b are summarised in Table 12. From this preliminary analysis, chloroform is the best candidate LLE solvent for CPMa due to its high extraction performance and material efficiency. However, in order to gain a full understanding of the performance of each separation option, cost estimates as well as assessment of extraction performance and material efficiencies are required.


Page 22 of 39

Page 23 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


T = 20 °C

T = 40 °C

T = 30 °C

(a)

(b)

(c)

(d)

DCM

T = 20 °C

(e)

TCM

T = 20 °C

(f)

nHex

nHep

nHex

T = 20 °C

(g)

CyHex

MeCyHex

NEAT

DCM

TCM

E (-)

AQUEOUS (waste) ORGANIC (product)

Figure 8: Performances of different LLE solvents and varying solvent-to-feed ratios and temperatures. CPMa (a) n-Hexane (b) n-Heptane (c) Cyclohexane (d) Methylcyclohexane (e) Dichloromethane (f) Chloroform; CPMb (g) n-Hexane + membrane.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 24 of 39

S:F

S:F CPM with carrier solvent nHex

nHep

CyHex

MeCyHex

DCM

TCM

CPM without carrier solvent nHex

Figure 9: E-factors and mass productivities for all solvents at various operating conditions. Table 12: Optimum extraction performances of all candidate LLE solvents API recovery (%) Separation T S:F Organic Aqueous (°C) (—) (product) (waste) Route Solvent n-Hexane 20 5 60.6 39.4 n-Heptane 20 5 53.1 46.9 Cyclohexane 20 5 88.3 11.7 CPMa Methylcyclohexane 20 5 81.1 18.9 Dichloromethane 20 0.5 98.5 1.5 Chloroform 20 0.5 98.7 1.3 CPMb n-Hexane 20 5 51.9 49.1


E (—) 40.31 48.15 27.18 31.06 3.59 3.43 31.55

MP (%) 2.42 2.03 15.91 9.44 21.79 48.73 6.71

Page 25 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


3.5 Economic Analysis For the purposes of the economic analysis conducted here, material requirements of the CPM process with continuous and batch separation trains have been increased to account for the calculated inefficiencies of the separations discussed previously. A comparison of different separation scenarios for both solvent and neat cases is shown in Figure 10 (see also Table 13). The FOB cost of all major units (as well as solenoid pumps used for metering of reagents, workup agents and solvents) has been considered explicitly, with computation of CapEx contributions. Operating (OpEx) pumping contributions are considered within Utilities (U/W), without explicitly comparing pumping requirements between batch and flow reactors, as per our published established methodology.41 Total cost savings are achieved for all continuous separations (CPMa/b) relative to the process with a batch separation (BX). This is due to the significant CapEx savings available by significant reductions in separation equipment requirements relative to the batch separation. The greatest total cost savings are attainable by implementing CPMb; this is due to the reduced material requirements which give significant CapEx savings from the reduction in equipment size in the absence of carrier solvent. Indeed, Table 7 shows that the highest contribution to CapEx (FOB) is R-101. In CPMa, the greatest CapEx savings are available with chloroform as LLE solvent. This is also due to the reduced material requirements compared to separations implementing other LLE solvents, accessible by the high extraction efficiency of both dichloromethane and chloroform. The greatest OpEx savings are available by using chloroform as LLE solvent in CPMa. This is due to the high extraction performance (i.e. material efficiency) and the low material cost compared to other solvents (Table 9). Indeed, the OpEx savings available with chloroform in CPMa are even greater than for CPMb due to the low cost of chloroform compared to n-hexane as well as the poor extraction performance of CPMb. OpEx savings are unavailable with n-hexane and cyclohexane in CPMa, due to the higher material costs of these solvents. The only processes which show an improvement in the E-factor are CPMa options using dichloromethane and chloroform. This is due to the high extraction performance of these solvents relative to the batch extraction with diethyl ether described by Snead and Jamison28. The focus of our technoeconomic evaluation methodology is to analyse and comparatively evaluate unit operation cost contributions to the process, in order to illustrate areas of potential cost advantages. Further comparisons of corresponding cost contributions between batch and CPM processes would require an even more detailed technoeconomic analysis, and hence necessitate the use of pilot plant operation data.



Table 13: Summary of costs (CapEx = GBP; OpEx = GBP y-1; Total = GBP) and savings (%) and material efficiencies (E-factor) for different separation trains. Separation Train BX CPMa Solvent Et2O n-C6H14 n-C7H16 CyHex MeCyHex DCM TCM S:F (-) 1.5 5 5 5 5 0.5 0.5 T (°C) 20 20 20 20 20 20 20 BLIC 234,340 180,945 (−22.8) 196,819 (−16.0) 147,908 (−36.9) 154,357 (−34.1) 140,412 (−40.1) 140,300 (−40.1) WCC 46,960 37,474 (−20.2) 40,062 (−14.7) 31,237 (−33.5) 31,300 (−33.4) 28,310 (−39.7) 28,286 (−39.8) 218,420 (−22.4) 236,881 (−15.8) 179,145 (−36.3) 185,657 (−34.0) 168,721 (−40.0) 168,586 (−40.1) CapEx 281,300

Total E (-)

(a)

3,674 (+39.9) 5,072 (+15.6) 8,746 (+24.7)

1,994 (−24.1) 3,833 (−12.7) 5,827 (−16.9)

4,730 (+80.1) 5,620 (+28.1) 10,351 (+47.6)

1,224 (−53.4) 2,416 (−45.0) 3,640 (−48.1)

649 (−75.3) 785 (−82.1) 1,434 (−79.6)

646 (−75.4) 774 (−82.4) 1,420 (−79.8)

2,436 (− 7.2) 3,237 (−26.2) 5,674 (−19.1)

368,726 10.21

327,415 (−11.2) 40.31

309,498 (−16.1) 48.15

308,144 (−16.4) 27.18

231,018 (−37.3) 31.06

186,593 (−49.4) 3.59

186,280 (−49.5) 3.43

128,533 (−65.1) 31.55

100%

CapEx OpEx Total

80% 60%

(b)

20% 0% -20% -40%

100%

WCC BLIC CapEx

80% 60%

Cost Savings (%)

40%

(c)

100%

40% 20% 0% -20% -40%

60% 40% 20% 0% -20% -40%

-60%

-60%

-80%

-80%

-80%

-100%

nHep CyHex MeCyHex DCM CPMa

TCM CPMb

U/W Materials OpEx

80%

-60%

nHex

CPMb n-C6H14 5 20 43,391 (−81.5) 8,763 (−81.3) 52,155 (−81.5)

2,626 4,389 7,015

Cost Savings (%)

Materials U/W OpEx

Cost Savings (%)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

Page 26 of 39

-100%

-100%

nHex

nHep CyHex MeCyHex DCM TCM CPMb CPMa

nHex

nHep CyHex MeCyHex DCM CPMa

Figure 10: Cost savings elements of different separation options relative to the batch separation (a) Total cost savings (b) CapEx savings (c) OpEx savings.


TCM

CPMb

Page 27 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


4.

Discussion

5.

Conclusions

The recently demonstrated continuous flow synthesis of diphenhydramine28 expedites the technoeconomic analysis of the upstream CPM of this popular, globally marketed drug. The demonstrated process features full atom economy and a single synthesis reaction which can be performed in either carrier solvent or a neat mixture. This work features the development of a process model of a continuous plant for CPM of 100 kg per annum of diphenhydramine via continuous flow synthesis in carrier solvent and as a neat mixture. By determination of kinetic parameters from published experimental data, reactor volumes of 18.45 mL and 3.22 mL have been calculated for solvent and neat syntheses, respectively. The obtained reactor volumes demonstrate the benefit of small equipment requirements available via CPM. Group contribution methods have been used for material property estimation. Such estimations have been required in order to obtain accurate calculations of reactor heating duties, which showed that considerable heating is required for both solvent and neat syntheses due to the elevated reactor operating temperatures28. Activity coefficient models have been required for calculation of API solubility in different solvents and for calculation of liquid-liquid equilibria (LLE). The partial lack of interaction parameters available for liquid-liquid equilibria necessitated the complementary use of UNIFAC as well as NRTL models in order to study different ternary systems in detail. Original thermodynamic studies are necessary for the provision of experimental ternary equilibrium data which would enable comparisons, corroboration and further validation of the model-based process simulation results and the technoeconomic conclusions presented. Several candidate solvents for continuous liquid-liquid extraction of API from the aqueous PFR effluent into an organic product phase have been considered based upon their propensity to form a multiphase mixture, favourable solubility of API in the candidate LLE solvent compared to the carrier solvent, and their inherent toxicity according to a Pfizer solvent selection guide74 and FDA guidelines. From an initial list of solvents, n-hexane, n-heptane, cyclohexane, methylcyclohexane, dichloromethane and chloroform have been selected as promising candidates for application. The evaluated solvents cover a wide range of toxicities and extraction performances. From a solely technical viewpoint, the best LLE solvent in terms of process performance and material efficiency (E-factor) appears to be chloroform: due to favourable ternary equilibrium thermodynamics, it demonstrates an impressively high API extraction fraction of 98.7%, which thus secures a miniscule E-factor of 3.43; this material efficiency appears appropriate for pharmaceutical processes, which are typically very wasteful. This solvent also promises the greatest total cost savings (-49.5%) and reduction in E-factor relative to the batch separation (E = 3.43 compared to E=10.21 for batch) of all separation cases for the solvent synthesis. Nevertheless, the known and extremely high toxicity of this solvent prohibits any recommendation for CPM consideration, because the subsequent solvent exchange (required for downstream processing and final formulation production) would be extremely costly, and any remnant traces would be catastrophic to product as well as patients. Implementing a neat CPM process (one without carrier solvent use for continuous flow synthesis) has also been considered, and allows even higher cost savings of all CPM processes considered (-65.1%). However, it is much less materially efficient than the solvent synthesis with chloroform as an extraction solvent (E=31.55). This is due to the inefficient extraction with n-hexane which is followed by a membrane separation. Methylcyclohexane is the next strongest performer among all candidate LLE solvents considered; it also presents remarkably advantageous characteristics (environmental factor of 31.06, total cost savings of 37.3%) and its lower toxicity renders it significantly more acceptable for CPM implementation. Indeed, the most environmentally friendly solvent analysed, n-heptane, shows the worst extraction performance and highest E-factor of all candidates considered. If greater amounts of solvent are acceptable for productionscale CPM use (i.e. S:F > 5), then greener solvents are an even more attractive option, which would though incur greater operating expenditures for the same API extraction level. This eloquent trade-off between high-performance (yet toxic) solvents and less advantageous LLE candidates (which are inherently safer due to significantly lower toxicity) is an undeniable reality underlining the importance and potential of model-based plantwide design, simulation and optimisation of CPM processes in pursuit of greener solvents.

Diphenhydramine is a promising candidate for Continuous Pharmaceutical Manufacturing. Its high global demand necessitates materially and economically efficient manufacturing techniques which have been explored via process modelling in this work. By illustrating the small reactor volumes typical of the CPM process, the material efficiencies available through continuous separation trains, and the associated



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

financial benefits, a strong case can be made for the advantages of this emerging technology over traditionally implemented, wasteful and expensive batch processes. This original technoeconomic analysis follows an established methodological approach, but for an API of high current academic and industrial interest (diphenhydramine). Several key assumptions affect our results (e.g. extent of solvent recovery, utility and waste handling unit costs, equipment vendor prices etc.), but they have been comprehensively documented to ensure reproducibility, and can easily be customised by an expert reader so as to study alternative conditions. The comprehensive evaluation of LLE solvents for a CPM of diphenhydramine implementation illustrates the importance of detailed technoeconomic analyses, and sensitivity analyses can be performed for specific parameters, in order to investigate if and how plantwide economic viability is affected by exogenous parametric uncertainty. The continuous flow synthesis of the API has been considered via two routes:28 with and without carrier solvent. The achievable conversions for both synthetic routes are quite similar, despite one route requiring no carrier solvent. As such, each route will have quite different material throughputs (indeed, this is one of the principal advantages in pursuing syntheses without carrier solvents), as our balances illustrate. Despite the small scale of diphenhydramine production considered in the study, the viability of CPM of diphenhydramine is evident28. The benefits of small equipment and material efficiency compared to batch processes are apparent for specific separation options. Furthermore, the benefits of implementing novel process options such as synthesising diphenhydramine in a neat mixture is shown in this study. Whilst the inefficiency of API extraction with this setup renders neat CPM a less economically viable option than the solvent CPM using chloroform for extraction, it highlights the material efficiencies, equipment reductions and cost savings available by implementing such novel process windows. Further experimental investigation of neat CPM processes at pilot plant and production scale should be conducted (with perhaps even more efficient separations than those explored here), in order to gain further, deeper understanding of these desirable operating conditions. It is also important that downstream processing options be considered in the model, to quantify the benefits of a fully integrated CPM process over current batch techniques. We have opted to maintain the reference currency used in our studies (UK pound sterling, GBP), not only to ensure a common basis and facilitate comparisons to our recent publications, but also in order to avoid confusion due to currency exchange rate variations in this period (2014-2017).

6.

Acknowledgements

The authors gratefully acknowledge the financial support of the Engineering and Physical Sciences Research Council (EPSRC) via a Doctoral Training Partnership (DTP) studentship to Mr S. Diab (Grant Number: EP/N509644/1). Tabulated and cited literature data suffice for reproduction of all original process simulation results, and no other supporting research data are stored or required.

7.

Nomenclature

Latin letters ai ais aij bi Ci Cj Cj,0 ci C1 C2 di E f Gi Gij hf0

Heat capacity group contribution estimation parameter for functional group i Solid component activity Interaction energy between components i and j in the NRTL model (J mol-1) Heat capacity group contribution estimation parameter for functional group i Group additivity coefficient for for group i in estimation of total phase change entropy Concentration of molecule j (mol L-1 or M) Initial concentration of molecule j (mol L-1 or M) Heat capacity group contribution estimation parameter for functional group i Cost of reference equipment (GBP) Scaled cost of designed equipment (GBP) Heat capacity group contribution estimation parameter for functional group i Environmental-factor Correction factor for cost capacity correlation Contribution of functional group i to total phase change entropy Coefficient of NRTL equation between components i and j in the NRTL model. Constant in estimating the standard enthalpy of formation ∆0 Hf (= 10 kJ mol-1)


Page 28 of 39

Page 29 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


h1i h2j ID k1 L Li Mj mAPI mbpd mur mus muAPI mwaste MP Ni n nCH2 ni NPV Qk qi R Rk r rA ri Spc T T0 Tfus Trxn t Umn XA Xf xi xm xisat z

Greek letters αij β Γk Γik γi γci γRi γsat i

Group contribution of first-order group i in estimation of standard enthalpy of formation Group contribution of second-order group j in estimation of standard enthalpy of formation Internal diameter (mm) First-order reaction rate constant (L mol-1 h-1) Reactor length (mm) UNIFAC compound parameter of r, q and z for molecule i Number of second-order functional groups Mass of recovered API (g h-1) Mass of byproducts (g h-1) Mass of unreacted reagents (g h-1) Mass of unrecovered solvent (g h-1) Mass of unrecovered API (g h-1) Mass of waste (g h-1) Mass productivity (%) Number of first-order functional groups Exponent particular to a particular piece of equipment Number of consecutive CH2 groups Number of functional groups Net Present Value (GBP) UNIFAC surface area parameter for functional group k UNIFAC parameter of molecule i, representing van der Waals molecular surface area Universal gas constant (=8.314 J mol-1K-1) UNIFAC volume parameter for functional group k Rate of interest (%) Rate of reaction of molecule A (mol L-1h-1) UNIFAC parameter of molecule i, representing van der Waals volume Total phase change entropy (J mol-1K-1) Absolute temperature (K) Standard temperature (=298.15 K) Melting point temperature (K) Reaction temperature (K) Time (h) UNIFAC energy of interaction between groups m and n Conversion of component A at time t Final conversion Mole fraction of component i UNIFAC mole fraction of group m Mole fraction of component i at saturation (solubility of component i) UNIFAC coordination number (=10)

Non-randomness parameter between components i and j in the NRTL model Coefficient of thermal expansion (K-1) UNIFAC residual group activity coefficient for group k UNIFAC residual group activity coefficient for group k in a solution of pure i UNIFAC activity coefficient of molecule i Combinatorial component of UNIFAC activity coefficient of molecule i Residual component of UNIFAC activity coefficient of molecule i UNIFAC activity coefficient of molecule i at saturation



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

∆0 Hf ∆Hfus ∆0 Hrxn ∆Hrxn ∆Tfus 0 Spc θi θm νik τi τij ϕi ψmn

Standard enthalpy of formation (J mol-1) Enthalpy of fusion (J mol-1) Standard enthalpy of reaction (J mol-1) Enthalpy of reaction (J mol-1) Total phase change entropy (J mol-1K-1) UNIFAC molecular-weighted area fraction component for molecule i UNIFAC summation of area fraction of group m over all groups Number of occurrences of group k on molecule i Residence time in reactor i (h) Dimensionless interaction parameter between components i and j in the NRTL model UNIFAC molecular-weighted segment fractional component of molecule i Interaction parameter between groups m and n in the UNIFAC model.

References 1. 2.

3. 4. 5. 6.

7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17.

18. 19. 20. 21.

Plumb, K. Continuous Processing in the Pharmaceutical Industry - Changing the Mind Set. Chem. Eng. Res. Des. 83, 730–738 (2005). Gernaey, K. V, Cervera-Padrell, A. E. & Woodley, J. M. Development of continuous pharmaceutical production processes supported by process systems engineering methods and tools. Future Med. Chem. 4, 1371–1374 (2012). European Commission Press Release Database, EU Industrial R&D Scoreboard 2012 (2013), Available at: http://europa.eu/rapid/press-release_MEMO-12-948_en.htm . (Accessed: 5th March 2016) Morgan, S., Grootendorst, P., Lexchin, J., Cunningham, C. & Greyson, D. The Cost of Drug Development: A Systematic Review. Health Policy (New. York). 100, 4–17 (2011). Behr, A. et al. New Developments in Chemical Engineering for the Production of Drug Substances. Eng. Life Sci. 4, 15–24 (2004). Gerogiorgis, D. I. & Jolliffe, H. G. Continuous pharmaceutical process engineering and economics Investigating technical efficiency, environmental impact and economic viability. Chem. Today 33, 29– 32 (2015). Plouffe, P., Macchi, A. & Roberge, D. M. From Batch to Continuous Chemical Synthesis — A Toolbox Approach. Org. Process Res. Dev. 18, 1286–1294 (2014). Yadav, V. G. & Stephanopoulos, G. Metabolic engineering: the ultimate paradigm for continuous pharmaceutical manufacturing. ChemSusChem 7, 1847–1853 (2014). Ashe, R. From Batch to Continuous Processing. Chem. Eng. 119, 34–40 (2012). Aref, H. Stirring by chaotic advection. J. Fluid Mech. 143, 1–21 (1984). Ottino, J. M. & Wiggins, S. Introduction: mixing in microfluidics. Philos. Trans. R. Soc. London Ser. aMathematical Phys. Eng. Sci. 362, 923–935 (2004). Yoshida, J., Nagaki, A. & Yamada, D. Continuous flow synthesis. Drug Discov. Today. Technol. 10, e53-9 (2013). Baumann, M. & Baxendale, I. R. The Synthesis of Active Pharmaceutical Ingredients (APIs) Using Continuous Flow Chemistry. Beilstein J. Org. Chem. 11, 1194–1219 (2015). Porta, R., Benaglia, M. & Puglisi, A. Flow Chemistry: Recent Developments in the Synthesis of Pharmaceutical Products. Org. Process Res. Dev. 20, 2–25 (2016). Malet-Sanz, L. & Susanne, F. Continuous Flow Synthesis. A Pharma Perspective. J. Med. Chem. 55, 4062–4098 (2012). Lee, S. L. et al. Modernizing Pharmaceutical Manufacturing: From Batch to Continuous Production. J. Pharm. Innov. 10, 191–199 (2015). Gernaey, K. V, Cervera-Padrell, A. E. & Woodley, J. M. Development of Continuous Pharmaceutical Production Processes Supported by Process Systems Engineering Methods and Tools. Future Med. Chem. 4, 1371–4 (2012). Troup, G. M. & Georgakis, C. Process systems engineering tools in the pharmaceutical industry. Comput. Chem. Eng. 51, 157–171 (2013). Rogers, A. & Ierapetritou, M. Challenges and Opportunities in Pharmaceutical Manufacturing Modelling and Optimization. Comput. Aided Chem. Eng. 34, 144–149 (2014). Zhao, C. et al. Toward intelligent decision support for pharmaceutical product development. J. Pharm. Innov. 1, 23–35 (2006). Mascia, S. et al. End-to-end Continuous Manufacturing of Pharmaceuticals: Integrated Synthesis,


Page 30 of 39

Page 31 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


22. 23.

24. 25. 26.

27. 28.

29. 30.

31. 32. 33. 34. 35. 36. 37.

38. 39. 40. 41. 42. 43. 44. 45.

46.

47.

Purification, and Final Dosage Formation. Angew. Chemie-International Ed. 52, 12359–12363 (2013). Kupracz, L. & Kirschning, A. Multiple Organolithium Generation in the Continuous Flow Synthesis of Amitriptyline. Adv. Synth. Catal. 355, 3375–3380 (2013). Palmer, E. GSK commits to continuous processing. (2013). Available at: http://www.fiercepharmamanufacturing.com/story/gsk-commits-continuous-processing/2013-0219?utm_medium=nl&utm_source=internal. (Accessed: 5th March 2016) Kopetzki, D., Lévesque, F. & Seeberger, P. H. A Continuous-Flow Process for the Synthesis of Artemisinin. Chem. - A Eur. J. 19, 5450–5456 (2013). Gustafsson, T., Sörensen, H. & Pontén, F. Development of a Continuous Flow Scale-Up Approach of Reflux Inhibitor AZD6906. Org. Process Res. Dev. 16, 925–929 (2012). Kuehn, S. E. Janssen Embraces Continuous Manufacturing for Prezista. Pharmaceutical Manufacturing (2015). Available at: http://www.pharmamanufacturing.com/articles/2015/janssen-embracescontinuous-manufacturing-for-prezista/. (Accessed: 5th October 2016) FDA Approves Tablet Production on Janssen Continuous Manufacturing Line. PharmaTech.com April 12, (2016). Snead, D. R. & Jamison, T. F. End-to-End Continuous Flow Synthesis and Purification of Diphenhydramine Hydrochloride Featuring Atom Economy, In-Line Separation, and Flow of Molten Ammonium Salts. Chem. Sci. 4, 2822–2827 (2013). Adamo, A. et al. On-Demand Continuous-Flow Production of Pharmaceuticals in a Compact, Reconfigurable System. Science (80-. ). 352, 61–67 (2016). Pastre, J. C., Browne, D. L., O’Brien, M. & Ley, S. V. Scaling Up of Continuous Flow Processes with Gases Using a Tube-in-Tube Reactor: Inline Titrations and Fanetizole Synthesis with Ammonia. Org. Process Res. Dev. 17, 1183–1191 (2013). Ahmed-Omer, B. & Sanderson, A. J. Preparation of fluoxetine by multiple flow processing steps. Org. Biomol. Chem. 9, 3854–62 (2011). LaPorte, T. L. et al. Development and scale-up of three consecutive continuous reactions for production of 6-hydroxybuspirone. Org. Process Res. Dev. 12, 956–966 (2008). Bogdan, A. R., Poe, S. L., Kubis, D. C., Broadwater, S. J. & McQuade, D. T. The Continuous-Flow Synthesis of Ibuprofen. Angew. Chemie-International Ed. 48, 8547–8550 (2009). Snead, D. R. & Jamison, T. F. A Three-Minute Synthesis and Purification of Ibuprofen: Pushing the Limits of Continuous-Flow Processing. Angew. Chemie-International Ed. 54, 983–987 (2015). Hopkin, M. D. et al. A flow-based synthesis of Imatinib: the API of Gleevec. Chem. Commun. 46, 2450–2452 (2010). Zhang, P., Russell, M. G. & Jamison, T. F. Continuous Flow Total Synthesis of Rufinamide. Org. Process Res. Dev. 18, 1567–1570 (2014). Murray, P. R. D. et al. Continuous Flow-Processing of Organometallic Reagents Using an Advanced Peristaltic Pumping System and the Telescoped Flow Synthesis of ( E/Z )-Tamoxifen. Org. Process Res. Dev. 17, 1192–1208 (2013). Martin, A. D., Siamaki, A. R., Belecki, K. & Gupton, F. A Flow-Based synthesis of Telmisartan. J. Flow Chem. 5, 145–147 (2015). Fuse, S., Mifune, Y., Tanabe, N. & Takahashi, T. Continuous-flow synthesis of activated vitamin D3 and its analogues. Org. Biomol. Chem. 10, 5205 (2012). Teoh, S. K., Rathi, C. & Sharratt, P. Practical Assessment Methodology for Converting Fine Chemicals Processes from Batch to Continuous. Org. Process Res. Dev. 20, 414–431 (2015). Schaber, S. D. et al. Economic Analysis of Integrated Continuous and Batch Pharmaceutical Manufacturing: A Case Study. Ind. Eng. Chem. Res. 50, 10083–10092 (2011). Jolliffe, H. G. & Gerogiorgis, D. I. Process Modelling and Simulation for Continuous Pharmaceutical Manufacturing of Ibuprofen. Chem. Eng. Res. Des. 97, 175–191 (2015). Jolliffe, H. G. & Gerogiorgis, D. I. Process Modelling and Simulation for Continuous Pharmaceutical Manufacturing of Artemisinin. Chem. Eng. Res. Des. 112, 310–325 (2016). Kwon, J.S.I., Nayhouse, M., Orkoulas, G., and Christofides, P.D. Modeling and control of crystal shape in continuous protein crystallization. Chem. Eng. Sci. 107, 47–57 (2014). Yang, Y., Song, L., Gao, T. & Nagy, Z. K. Integrated Upstream and Downstream Application of Wet Milling with Continuous Mixed Suspension Mixed Product Removal Crystallization. Cryst. Growth Des. 15, 5879–5885 (2015). Sen, M., Chaudhury, A., Singh, R., John, J. & Ramachandran, R. Multi-scale flowsheet simulation of an integrated continuous purification–downstream pharmaceutical manufacturing process. Int. J. Pharm. 445, 29–38 (2013). Borsos, Á., Majumder, A. & Nagy, Z. K. in Computer Aided Chemical Engineering 33, 781–786 (2014).



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

48.

49. 50.

51. 52. 53. 54. 55. 56. 57. 58.

59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72.

73. 74. 75.

76.

Yang, Y. & Nagy, Z. K. Model-Based Systematic Design and Analysis Approach for Unseeded Combined Cooling and Antisolvent Crystallization (CCAC) Systems. Cryst. Growth Des. 14, 687–698 (2014). Nagy, Z. K. & Braatz, R. D. Advances and New Directions in Crystallization Control. Annu. Rev. Chem. Biomol. Eng. 3, 55–75 (2012). Su, Q., Benyahia, B., Nagy, Z. K. & Rielly, C. D. Mathematical Modeling, Design, and Optimization of a Multisegment Multiaddition Plug-Flow Crystallizer for Antisolvent Crystallizations. Org. Process Res. Dev. 19, 1859–1870 (2015). Nayhouse, M. et al. Modeling and control of ibuprofen crystal growth and size distribution. Chem. Eng. Sci. 134, 414–422 (2015). Benyahia, B., Lakerveld, R. & Barton, P. I. A Plant-Wide Dynamic Model of a Continuous Pharmaceutical Process. Ind. Eng. Chem. Res. 51, 15393–15412 (2012). Gerogiorgis, D. I. & Barton, P. I. Steady-state optimization of a continuous pharmaceutical process. 10th Int. Symp. Process Syst. Eng. 27, 927–932 (2009). Lakerveld, R. et al. The Application of an Automated Control Strategy for an Integrated Continuous Pharmaceutical Pilot Plant. Org. Process Res. Dev. 19, 1088–1100 (2015). Baxendale, I. R. et al. Achieving continuous manufacturing: technologies and approaches for synthesis, workup, and isolation of drug substance. J. Pharm. Sci. 104, 781–791 (2015). Jolliffe, H. G. & Gerogiorgis, D. I. Process Modelling and Simulation for Continuous Pharmaceutical Manufacturing of Ibuprofen. Chem. Eng. Res. Des. 97, 175–191 (2015). Kopetzki, D., Levesque, F. & Seeberger, P. H. A Continuous-Flow Process for the Synthesis of Artemisinin. Chem. Eur. J. 19, 5450–5456 (2013). Jolliffe, H. G. & Gerogiorgis, D. I. Plantwide Design and Evaluation of Two Continuous Pharmaceutical Manufacturing (CPM) Cases: Ibuprofen and Artemsinin. Comput. Aided Chem. Eng. (2016). doi:10.1016/B978-0-444-63576-1.50063-7 Richardson, G. S., Roehrs, T. A., Rosenthal, L., Koshorek, G. & Roth, T. Tolerance to daytime sedative effects of H1 antihistamines. J. Clin. Psychopharmacol. 22, 511–515 (2002). Ravina, E. & Kubinyi, H. The evolution of drug discovery. (2011). Rieveschl, G. Dialkylaminoalkyl benzhydryl ethers and salts thereof. 2121714, (1947). Anastas, P. T. & Warner, J. C. Green Chemistry: Theory and Practice. (Oxford University Press, 1998). Ferraz, R., Branco, L. C., Prudencio, C., Noronha, J. P. & Petrovski, Z. Ionic liquids as active pharmaceutical ingredients. ChemMedChem 6, 975–985 (2011). Olivier-Bourbigou, H., Magna, L. & Morvan, D. Ionic liquids and catalysis: Recent progress from knowledge to applications. Appl. Catal. a-General 373, 1–56 (2010). Renken, A. et al. Ionic liquid synthesis in a microstructured reactor for process intensification. Chem. Eng. Process. 46, 840–845 (2007). Waterkamp, D. A. et al. Synthesis of ionic liquids in micro-reactors - a process intensification study. Green Chem. 9, 1084–1090 (2007). Constantinou, L. & Gani, R. New group-contribution method for estimating properties of pure compounds. AIChE J. 40, 1697–1710 (1994). Rihani, D. N. & Doraiswamy, L. K. Estimation of heat capacity of organic compounds from group contributions. Ind. Eng. Chem. Fundam. 4, 17–21 (1965). Fredenslund, A., Jones, R. L. & Prausnitz, J. M. Group-Contribution Estimation of Activity-Coefficients in Nonideal Liquid-Mixtures. AIChE J. 21, 1086–1099 (1975). Abrams, D. S. & Prausnitz, J. M. Statistical thermodynamics of liquid-mixtures - new expression for excess Gibbs energy of partly or completely miscible systems. AIChE J. 21, 116–128 (1975). Chickos, J. S. & Acree, W. E. Estimating solid-liquid phase change enthalpies and entropies. J. Phys. Chem. Ref. Data 28, 1535–1673 (1999). Tripathi, B. S., Awasthi, A., Pandey, P. K. & Awasthi, A. Thermal expansion coefficient of ternary liquid mixture using hard sphere models and Flory’s statistical theory. Indian J. Phys. 84, 449–458 (2010). Gurvich, L. V et al. Thermodynamic properties of alkali metal hydroxides .1. Lithium and sodium hydroxides. J. Phys. Chem. Ref. Data 25, 1211–1276 (1996). Alfonsi, K. et al. Green Chemistry Tools to Influence a Medicinal Chemistry and Research Chemistry Based Organisation. Green Chem. 10, 31–36 (2008). FDA. Guidance for Industry Q3C - Tables and List. (2012). Available at: http://www.fda.gov/downloads/drugs/guidancecomplianceregulatoryinformation/guidances/ucm073395. pdf. (Accessed: 25th April 2016) Renon, H. & Prausnitz, J. M. Local Compositions in Thermodynamic Excess Functions for Liquid Mixtures. AIChE J. 14, 135–144 (1968).


Page 32 of 39

Page 33 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


77.

78.

79.

80. 81. 82. 83. 84. 85. 86.

87. 88. 89. 90. 91.

92.

93.

94. 95. 96.

97. 98. 99.

Ferreira, P. O., Ferreira, J. B. & Medina, A. G. Liquid liquid equilibria for the system Nmethylpyrrolidone + toluene + n-heptane - UNIFAC interaction parameters for N-methylpyrrolidone. Fluid Phase Equilib. 16, 369–379 (1984). Gebreyohannes, S., Neely, B. J. & Gasem, K. A. M. Generalized nonrandom two-liquid (NRTL) interaction model parameters for predicting liquid-liquid equilibrium behavior. Ind. Eng. Chem. Res. 53, 12445–12454 (2014). Diab, S. & Gerogiorgis, D. I. Process modelling and simulation for continuous pharmaceutical manufacturing of diphenhydramine. 2015 AIChE Annual Meeting, Salt Lake City, UT, USA. (2015). Available at: https://aiche.confex.com/aiche/2015/webprogram/Paper431784.html. (Accessed: 2nd February 2016) Chen, J.-T. & Chen, M.-C. Salting effect on the liquid–liquid equilibria for the ternary system water+Nmethyl-2-pyrrolidone+1-pentanol. Fluid Phase Equilib. 266, 1–7 (2008). Chen, J.-T. & Chen, M.-C. Liquid–liquid equilibria for the quaternary systems of sater + N -Methyl-2pyrrolidone + 1-Hexanol + NaCl, + KCl, or + KAc. J. Chem. Eng. Data 53, 217–222 (2008). Hughes, J. P. Toxicity of carbon tetrachloride. Br. Med. J. 3, 360 (1969). Constable, D. J. C., Curzons, A. D. & Cunningham, V. L. Metrics to green chemistry - which are the best? Green Chem. 4, 521–527 (2002). Sheldon, R. A. Fundamentals of green chemistry: efficiency in reaction design. Chem. Soc. Rev. 41, 1437–1451 (2012). Ritter, S. K. Reducing Environmental Impact of Organic Synthesis. Chem. Eng. News 91, 22–23 (2013). Curzons, A. D., Mortimer, D. N., Constable, D. J. C. & Cunningham, V. L. So you think your process is green, how do you know? Using principles of sustainability to determine what is green – a corporate perspective. Green Chem. 3, 1–6 (2001). Sorensen, J. M. & Arlt, W. Liquid-liquid Equilibrium Data Collection (Ternary systems). DECHEMA Vol. V, Part 2-3 (1981). Available at: http://detherm.cds.rsc.org/. Du, S. et al. Correlation and thermodynamic analysis of solubility of diphenhydramine hydrochloride in pure and binary solvents. J. Chem. Thermodyn. 93, 132–142 (2015). Woods, D. R. Rules of Thumb in Engineering Practice. (Wiley, 2007). Couper, J. R. Process Engineering Economics. (CRC Press, 2003). Corning. Corning Advanced-Flow G1 SiC Reactor. (2015). Available at: http://www.corning.com/media/worldwide/global/documents/G1_SiC_leaflet_FINAL_6.1.15.pdf. (Accessed: 8th March 2016) Cole-Parmer. Cole Parmer Polystat Advanced 15L Heat Cool Bath 35 to 200C 115VAC from ColeParmer United Kingdom. (2015). Available at: http://www.coleparmer.co.uk/Product/Cole_Parmer_Polystat_Advanced_15L_Heat_Cool_Bath_35_to_ 200C_115VAC/WZ-12122-56. (Accessed: 8th March 2016) ProMinent. Solenoid Driven Metering Pumps. (2015). Available at: https://www.prominent.co.uk/en/Products/Products/Metering-Pumps/Solenoid-Driven-MeteringPumps/pg-solenoid-driven-metering-pumps.html. (Accessed: 8th March 2016) Adamo, A., Heider, P. L., Weeranoppanant, N. & Jensen, K. F. Membrane-based, liquid-liquid separator with integrated pressure control. Ind. Eng. Chem. Res. 52, 10802–10808 (2013). Calabrese, G. S. & Pissavini, S. From batch to continuous flow processing in chemicals manufacturing. AIChE J. 57, 828–834 (2011). Maczynski, A. et al. IUPAC-NIST Solubility Data Series. 81. Hydrocarbons with water and seawaterrevised and updated. Part 5. C-7 hydrocarbons with water and heavy water. J. Phys. Chem. Ref. Data 34, 1399–1487 (2005). Magnussen, T., Rasmussen, P. & Fredenslund, A. UNIFAC Parameter Table for Prediction of LiquidLiquid Equilibria. Ind. Eng. Chem. Process Des. Dev. 20, 331–339 (1981). Wittig, R., Lohmann, J. & Gmehling, J. Vapor-Liquid Equilibria by UNIFAC Group Contribution. 6. Revision and Extension. Ind. Eng. Chem. Res. 42, 183–188 (2003). Gracin, S., Brinck, T. & Rasmuson, A. C. Prediction of Solubility of Solid Organic Compounds in Solvents by UNIFAC. Ind. Eng. Chem. Res. 41, 5114–5124 (2002).

Appendix A: Reaction Kinetics and Reactor Design The differential forms of the rate equations for first and second-order reactions are shown in A1 and A2, respectively. A2 holds only when the initial concentrations of reactants A and B in the second-order reaction are equal, as in the neat reaction.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

−

dCA = k1 CA dt

−

dCA = k1C2A dt

Page 34 of 39

(A1) (A2)

By integration of A1 and A2, the following are obtained for first- and second-order reactions respectively: ln CA = ln CA0 − k1 t 1 1 = k2 t + CA CA0

(A3) (A4)

From A3, plotting ln CA versus time results in a linear correlation with a slope of minus k1. In A4, 1 plotting C versus time results in a linear correlation with a slope of k2. A

The rate equations for the solvent and neat reactions are given by A5 and A6 respectively: −rCDPM = k1 CCDPM −rCDPM = k2 C2CDPM

(A5) (A6)

A5 and A6 can be expressed in terms of conversion of CDPM by A7 and A8 respectively: −rCDPM = k1CA0 (1 - XA)

−rCDPM =

k2C2A0(1 - XA)2

(A7) (A8)

For the solvent reaction, the integral form of the PFR design equation is used to calculate the required residence time: XCDPM,f

1 τi = k1 0

dXCDPM (1 - XCDPM)

(A9)

For the neat reaction, the integral form of the PFR design equation can be used again: XCDPM,f

1 τi = k2 CA0 0

dXCDPM (1 - XCDPM)2

(A10)

Appendix B: Entropy and Enthalpy of Fusion The total phase change entropy is approximated as the entropy of fusion and is estimated using Equation A11 for aliphatic and benzenoid aromatic hydrocarbons: ∆0 fus Spc = ni Gi + nCH2 CCH2 GCH2

(A11)

1.31, nCH2 ≥ ni , i ≠ CH2 CCH2 = i 1.00, otherwise

(A12)

T

i

Gi represents the contribution of group i to the total phase change entropy. Group coefficients ni, Gi and Ci are used for multiple occurrences of group i on a molecule. The calculation requires tabulated values for Gi and Ci from the literature 71.


Page 35 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


The enthalpy of fusion can then be calculated from the estimated entropy of fusion as follows: ∆Hfus ≅ ∆0 fus Sps Tfus T

(A13)

Appendix C: UNIFAC Model The UNIFAC activity coefficient for molecule i (γi) is the sum of a combinatorial and residual component 69 : ln γi = ln γci + ln γri

(A14)

The combinatorial component is calculated via the UNIQUAC model 70: ln γci = ln

ϕi z ϕ θi + q ln + Li − i xj Lj xi 2 i ϕi xi

(A15)

j

φi and θi are molar weighted segment and area fractional components respectively. Li is a UNIFAC compound parameter defined by parameters ri, qi and z (here, z=10). ϕi = θi =

xi ri ∑j xj rj

(A16)

xi qi ∑j xj qj

(A17)

z Li = ri − qi − (ri − 1) 2

(A18)

ri and qi are calculated from contributions of volume and surface area parameters for each functional group (R and Q, respectively) weighted by their occurrence on each a molecule (vki): ri = vik Rk

(A19)

k

qi = vik Qk

(A20)

k

The residual component of the activity coefficient is calculated by Equation A21: ln γri = vik ln Γk − ln Γik

(A21)

k

Γk and Γki are residual group activity coefficients of group k in reality and in a reference solution of pure substance I, respectively. Both are calculated as follows: ln Γk = Qk 1 - ln θm ψmk − m

m

θm ψkm ∑n θm ψnm

(A22)

θm is the summation of the area fraction of group m over all different groups. θm = xm is the mole fraction of group m:

Qm x m ∑n Qn xn


(A23)


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

xm

j ∑j v m xj

∑j ∑n vnjxj

Page 36 of 39

(A24)

ψmn represents the interaction between different functional groups in solution, and is calculated using an Arrhenius-type expression. Tabulated values of R and Q are required for calculation of the UNIFAC activity coefficient of molecule i, and are available in the literature 98. ψmn = exp −

Umn − Unm RT

(A25)

Appendix D: NRTL Model The NRTL activity coefficient model for a multicomponent system is described by: ln γi =

n ∑j=1 xj τji Gji

∑nk=1 xk Gki

+ n

j=1

∑nm=1 xm τmj Gmj xk Gij τ − ij ∑nk=1 xk Gkj ∑nk=1 xkGkj

Gij = exp-αij τij

(A26)

(A27)

Here, αij is the non-randomness parameter between components i and j. The dimensionless interaction parameter, τij, gives the temperature dependence of the model. τij =

aij RT

(A28)

aij is the interaction between components i and j, which can be found in the literature 78.

Appendix E: Solubility Calculation The solid solubility of component i in mixture (xisat) is calculated as follows 99: ∆Hfus 1 1 sat asi = xsat − i γi = exp R Tfus T

(A29)

The enthalpy of fusion is calculated from the estimation method described by A11-A13 and the activity coefficient at saturation is calculated using A15. Iteration is required to calculate the solubility of solute i, as the activity coefficient is a function of mole fraction.

Appendix F: Purification Design Results Table A1 and A2 show the values illustrated in Figure 8 for all candidate LLE solvents at the various solvent-to-feed ratios considered.


Page 37 of 39

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48


Table A1: Evaluation of candidate LLE solvents at various solvent-to-feed ratios and temperatures. S:F (-) Solvent flow (g h-1) T (⁰C) Solvent Phase (O/A) API (g h-1) n-C6H14 API recovery (%) E (-) API (g h-1) n-C7H16 API recovery (%) E (-) API (g h-1) CyHex API recovery (%) E (-) API (g h-1) MeCyHex API recovery (%) E (-) API (g h-1) DCM API recovery (%) E (-) API (g h-1) TCM API recovery (%) E (-) S:F (-) Solvent flow (g h-1) T (⁰C) Solvent Phase (O/A) API (g h-1) n-C6H14 API recovery (%) E (-) API (g h-1) n-C7H16 API recovery (%) E (-) API (g h-1) CyHex API recovery (%) E (-) API (g h-1) MeCyHex API recovery (%) E (-) API (g h-1) DCM API recovery (%) E (-) API (g h-1) TCM API recovery (%) E (-)

0.25 20.99 30

20 O A 2.69 15.31 85.04 14.96 13.89 3.05 14.95 83.04 16.96 15.49 6.30 11.70 65.01 34.99 5.28 4.78 13.22 73.44 26.56 9.60

O A 2.66 15.34 85.19 14.81 14.18 2.39 15.61 86.71 13.29 20.03 3.22 14.78 82.13 17.87 11.30 3.77 14.23 79.07 20.93 12.46

40 O A 2.00 16.00 88.91 11.09 19.05 2.24 15.76 87.54 12.46 21.45 4.32 13.68 76.00 24.00 8.06 3.54 14.46 80.31 19.69 12.45

Fully Miscible

n/a

n/a

n/a

n/a

17.64 97.99

0.36 2.01 1.05

20 O A 8.83 9.17 50.94 49.06 20.79 8.12 9.88 54.89 45.11 25.30 13.49 4.51 25.04 74.96 13.16 12.51 5.49 30.52 69.48 16.11 17.89 0.11 0.62 99.38 11.26 17.89 0.11 0.63 99.37 10.95

2 167.96 30 O

A

8.64 47.99 21.35 6.36 35.35 32.57 12.38 68.77 14.51 9.80 54.46 20.74

9.36 52.01 11.64 64.65 5.62 31.23 8.20 45.54

40 O A 6.57 11.43 63.47 36.53 28.28 5.96 12.04 66.89 33.11 34.84 12.36 5.64 31.36 68.64 14.45 9.29 8.71 48.37 51.63 19.78

n/a

n/a

n/a

n/a

20 O A 4.47 13.53 75.15 24.85 12.84 4.88 13.12 72.90 27.10 15.17 8.39 9.61 53.40 46.60 6.29 7.55 10.45 58.06 41.94 9.47 17.74 0.26 1.47 98.53 3.59 17.77 0.23 1.30 98.70 3.43

20 O A 9.89 8.11 45.03 54.97 27.21 8.81 9.19 51.08 48.92 32.86 14.72 3.28 18.23 81.77 17.83 13.54 4.46 24.77 75.23 21.05 17.92 0.08 0.47 99.53 15.79 17.92 0.08 0.45 99.55 15.74

0.5 41.99 30 O A 4.39 13.61 75.62 24.38 13.18 3.81 14.19 78.83 21.17 19.65 5.88 12.12 67.36 32.64 9.42 5.95 12.05 66.95 33.05 12.29

40 O A 3.34 14.66 81.45 18.55 17.56 3.58 14.42 80.09 19.91 21.02 6.94 11.06 61.42 38.58 7.77 5.62 12.38 68.81 31.19 12.10

n/a

n/a

n/a

n/a

3 251.94 30 O

A

9.68 53.77 27.95 6.90 38.33 42.22 13.89 77.14 19.00 10.67 59.25 27.00

8.32 46.23 11.10 61.67 4.11 22.86 7.33 40.75

40 O A 7.33 10.67 59.25 40.75 37.04 6.46 11.54 64.10 35.90 45.15 13.54 4.46 24.78 75.22 19.47 10.06 7.94 44.11 55.89 25.59

n/a

n/a

n/a

n/a


20 O A 6.66 11.34 62.97 37.03 14.81 6.69 11.31 62.81 37.19 18.03 10.97 7.03 39.07 60.93 8.52 10.34 7.66 42.53 57.47 11.36 17.79 0.21 1.16 98.84 6.16 17.83 0.17 0.96 99.04 6.12

20 O A 10.52 7.48 41.57 58.43 33.74 9.33 8.67 48.17 51.83 40.46 15.43 2.57 14.27 85.73 22.50 14.19 3.81 21.16 78.84 26.04 17.94 0.06 0.33 99.67 20.78 17.94 0.06 0.32 99.68 20.65

1 83.98 30 O A 6.55 11.45 63.61 36.39 15.21 5.25 12.75 70.85 29.15 23.29 9.17 8.83 49.03 50.97 10.44 8.15 9.85 54.72 45.28 14.68

40 O A 4.98 13.02 72.33 27.67 20.21 4.92 13.08 72.69 27.31 24.93 9.75 8.25 45.81 54.19 9.67 7.69 10.31 57.27 42.73 14.24

n/a

n/a

n/a

n/a

4 335.92 30 O 10.27 57.07 34.65 7.27 40.41 51.94 14.77 82.05 23.58 11.18 62.09 33.34

40 A 7.73 42.93 10.73 59.59 3.23 17.95 6.82 37.91

O 7.79 43.29 45.87 6.78 37.66 55.55 14.18 78.78 24.59 10.54 58.57 31.48

n/a

n/a

n/a

n/a

A 10.21 56.71 11.22 62.34 3.82 21.22 7.46 41.43


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

Solvent-to-feed ratio Solvent flow (g h-1) T (⁰C) Solvent Phase (O/A) API content (g h-1) n-C6H14 API recovery (%) E (-) API content (g h-1) n-C7H16 API recovery (%) E (-) API content (g h-1) CyHex API recovery (%) E (-) API content (g h-1) MeCyHex API recovery (%) E (-) API content (g h-1) DCM API recovery (%) E (-) API content (g h-1) TCM API recovery (%) E (-)

Page 38 of 39

5 419.9 30

20 O 10.91 60.58

A 7.09 39.42

O 10.62 59.02

8.45 46.94

7.48 41.58

2.11 11.70

15.33 85.18

3.41 18.94

11.38 63.20

40.31

40 A 7.38 40.98

O 8.09 44.93

10.52 58.42

7.01 38.94

2.67 14.82

14.59 81.03

6.62 36.80

10.84 60.22

A 9.91 55.07

41.38

9.55 53.06 48.15

54.70 10.99 61.06

61.74

15.89 88.30 27.18

65.99 3.41 18.97

28.20

14.59 81.06 31.06 17.96 99.76

29.72 7.16 39.78

39.72

37.39

0.04 0.24

n/a

n/a

0.04 0.23

n/a

n/a

25.24 17.96 99.77 25.30

Table A2: Evaluation of neat separation (CPMb) at various solvent-to-feed ratios at T = 20 °C. Solvent-to-feed ratio Solvent flow (g h-1) Solvent Phase (O/A) API content (g h-1) n-C6H14 API recovery (%) E (-)

0.25 14.24 O 2.23 12.38

0.5 28.48 A 15.77 87.62

13.90

O 3.72 20.66

1 56.96 A 14.28 79.34

11.76

O 5.59 31.05

2 113.92 A 12.41 68.95

12.60

O 7.46 41.47

3 170.88 A 10.54 58.53

16.82


O 8.40 46.69

4 227.84 A 9.60 53.31

21.62

O 8.97 49.83

5 284.8 A 9.03 50.17

26.55

O 9.35 51.92

A 8.65 48.08 31.55

[ TABLE OF CONTENTS GRAPHIC ]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47


Cost Comparison (CM vs. BX)

Page 39 of 39


Process Modeling, Simulation, and Technoeconomic Evaluation of

Recommend Documents