# End-to-End Continuous Manufacturing (E2E-CM) in API Plants: Integrated Flow Synthesis, Crystallization, Filtration & Drying

# Executive Summary & Industrial Context

In traditional batch pharmaceutical manufacturing, the synthesis of an Active Pharmaceutical Ingredient (API) involves discrete unit operations—batch reaction, liquid-liquid extraction, batch crystallization, nutsche filtration, tray drying, and dry milling—separated by hold times, intermediate testing, and off-line quality assurance. This disconnected paradigm results in extended cycle times (often 30–90 days per batch), high inventory holding costs, large plant footprints, and significant risk of batch loss due to non-uniform heat and mass transfer.

End-to-End Continuous Manufacturing (E2E-CM) integrates flow synthesis, workup, crystallization, solid-liquid separation, and drying into a single, automated, uninterrupted continuous process train. By operating at steady state, E2E-CM achieves dramatic process intensification: plant footprint is reduced by up to $70\%$ , solvent inventory by $80\%$ , and manufacturing lead times from months to hours.

This guide provides a comprehensive chemical engineering roadmap for designing, sizing, controlling, and validating an integrated E2E-CM API manufacturing facility in compliance with ICH Q13 guidelines.

# 1. Integrated Process Flow & Plant Architecture

An end-to-end continuous API manufacturing plant comprises five tightly integrated continuous unit operations:

                            END-TO-END CONTINUOUS API PLANT FLOWSHEET
 ┌─────────────────────────────────────────────────────────────────────────────────────────────┐
 │                                                                                             │
 │  ┌──────────────┐     ┌──────────────┐     ┌──────────────┐     ┌──────────────┐            │
 │  │ 1. REACTION  │────►│ 2. WORKUP &  │────►│ 3. MSMPR     │────►│ 4. ROTARY    │────┐       │
 │  │    SKID      │     │    EXTRACTION│     │CRYSTALLIZER  │     │   FILTRATION │    │       │
 │  │ (PFR / CSTR) │     │ (Membrane/   │     │ (Cooling/    │     │   & WASHING  │    │       │
 │  │              │     │  Centrifugal)│     │ Antisolvent) │     │              │    │       │
 │  └──────────────┘     └──────────────┘     └──────────────┘     └──────────────┘    │       │
 │                                                                                     ▼       │
 │  ┌──────────────┐     ┌──────────────┐     ┌──────────────┐                  ┌──────────────┐
 │  │ 6. FINAL     │◄────│ 5. CONTINUOUS│◄────│ OOS DIVERTER │◄─────────────────│ INLINE PAT   │
 │  │    PACKAGING │     │    DRYER     │     │    VALVE     │  (Reject Stream) │ (Raman/NIR/  │
 │  │    (Drums)   │     │ (Paddle/OBC) │     │  (3-Way)     │                  │  Particle)   │
 │  └──────────────┘     └──────────────┘     └──────────────┘                  └──────────────┘
 └─────────────────────────────────────────────────────────────────────────────────────────────┘

# 2. Unit Operation 1: Continuous Flow Reaction Skid Design

# 2.1 Plug Flow Reactor (PFR) vs. CSTR Cascade Hydraulics

The reaction module is designed based on intrinsic chemical kinetics. For fast, highly exothermic reactions ( $k > 1.0\,\text{L/mol}\cdot\text{s}$ ), a Plug Flow Reactor (PFR) with high surface-area-to-volume ratio ( $A/V > 3,000\,\text{m}^2/\text{m}^3$ ) is specified. For slower reactions requiring solid suspensions, a cascade of Continuous Stirred Tank Reactors (CSTRs) is deployed.

For $N$ identical CSTRs in series, the fractional conversion $X_N$ for a first-order reaction ( $r = k C$ ) is governed by:

C_N = \frac{C_0}{(1 + k \tau_i)^N}

where $\tau_i = V_i / Q_{total}$ is the residence time of an individual CSTR stage.

                  CSTR CASCADE CONVERSION VS. STAGE COUNT (N)
 ┌─────────────────────────┬─────────────────────────┬─────────────────────────┐
 │ Stage Count ( $N$ )       │ Individual Stage  $\tau_i$ │ Overall Conversion ( $X$ )│
 ├─────────────────────────┼─────────────────────────┼─────────────────────────┤
 │ 1 CSTR                  │  $30\,\text{min}$         │  $75.0\%$                 │
 │ 2 CSTRs in series       │  $15\,\text{min}$         │  $88.9\%$                 │
 │ 3 CSTRs in series       │  $10\,\text{min}$         │  $93.8\%$                 │
 │ 5 CSTRs in series       │  $6\,\text{min}$          │  $97.5\%$                 │
 │ 1 PFR (Ideal Plug Flow) │  $30\,\text{min}$  (Total)│  $99.2\%$                 │
 └─────────────────────────┴─────────────────────────┴─────────────────────────┘

# 3. Unit Operation 2: Continuous Workup & Liquid-Liquid Extraction

Continuous extraction replaces batch phase splits using membrane-based phase separators (e.g., Zaiput Flow Technologies) or multi-stage centrifugal contactors (e.g., CINC).

# 3.1 Membrane Phase Separation Dynamics

Membrane separators utilize a hydrophobic PTFE membrane with pore diameter $d_p = 0.5\,\mu\text{m}$ . The separation is driven by differential capillary pressure ( $\Delta P_{\text{cap}}$ ):

\Delta P_{\text{cap}} = \frac{2 \gamma \cos\theta}{r_{pore}}

where $\gamma$ is interfacial tension, $\theta$ is contact angle, and $r_{pore}$ is pore radius.

To achieve $100\%$ aqueous-organic separation without phase breakthrough, the applied system pressure differential must be strictly maintained within:

\Delta P_{\text{transmembrane}} < \Delta P_{\text{cap}}

# 4. Unit Operation 3: Continuous Crystallization (MSMPR & OBC)

Crystallization determines key Quality Attributes (CQAs) of the API: Particle Size Distribution (PSD), polymorphism, and purity.

# 4.1 Mixed Suspension Mixed Product Removal (MSMPR) Design

An MSMPR crystallizer operates at steady state with continuous feed and product withdrawal. The population balance equation (PBE) for crystal length $L$ assuming size-independent growth rate $G$ and zero breakage/aggregation is:

\frac{\partial n(L)}{\partial t} + G \frac{\partial n(L)}{\partial L} + \frac{n(L)}{\tau} = 0

At steady state ( $\partial n / \partial t = 0$ ), the population density $n(L)$ follows an exponential distribution:

n(L) = n_0 \exp\left( -\frac{L}{G \tau} \right)

where $n_0 = B_0 / G$ is the nuclei density, and $B_0$ is the primary/secondary nucleation rate:

B_0 = k_b \cdot (\Delta C)^b \cdot M_T^j

Here $\Delta C = C - C^*$ is supersaturation, and $M_T$ is magma density ( $\text{kg/m}^3$ ).

                 MSMPR CRYSTALLIZER POPULATION DENSITY LOG-PLOT
  ln(n)
    ▲
 n0 ┼───\
    │    \
    │     \  Slope = -1 / (G * τ)
    │      \
    │       \
    └────────\─────────────────────► Crystal Size L (μm)

# 4.2 Multi-Stage MSMPR Cascade for Narrow PSD

A single MSMPR produces a wide exponential PSD with a high coefficient of variation ( $CV = 100\%$ ). Cascading 3 to 4 MSMPR stages in series narrows the PSD significantly ( $CV < 45\%$ ) while enabling staged temperature or antisolvent dosing profiles to maximize yield without causing oiling-out.

# 5. Unit Operation 4: Continuous Solid-Liquid Filtration & Washing

Continuous filtration is executed using a Continuous Rotary Drum Vacuum Filter (RDVF) or a Carousel Vacuum Filter.

# 5.1 Cake Filtration Kinetics & Darcy's Law

The rate of continuous cake filtration is governed by Darcy's equation modified for cake buildup:

\frac{dV}{dt} = \frac{A^2 \cdot \Delta P}{\mu \left( \alpha \cdot w \cdot V + R_m \cdot A \right)}

Integrating for constant pressure drop ( $\Delta P$ ):

\frac{t}{(V/A)} = \frac{\mu \cdot \alpha \cdot w}{2 \Delta P} \left(\frac{V}{A}\right) + \frac{\mu \cdot R_m}{\Delta P}

where:

$\alpha$ is specific cake resistance ( $\text{m/kg}$ ),
$w$ is dry cake mass per unit filtrate volume ( $\text{kg/m}^3$ ),
$R_m$ is filter medium resistance ( $\text{m}^{-1}$ ),
$\mu$ is liquid viscosity ( $\text{Pa}\cdot\text{s}$ ).

# 5.2 Rotary Drum Filtration Sizing Table

                  ROTARY DRUM VACUUM FILTER (RDVF) DESIGN PARAMS
 ┌───────────────────────────────────┬───────────────────┬───────────────────┐
 │ Process Parameter                 │ Design Value      │ Engineering Units │
 ├───────────────────────────────────┼───────────────────┼───────────────────┤
 │ Slurry Feed Rate                  │ 120               │ L/h               │
 │ Solid Content in Slurry           │ 15.0              │ wt%               │
 │ Specific Cake Resistance ( $\alpha$ )│  $2.4 \times 10^{11}$ │ m/kg             │
 │ Filter Area Required ( $A$ )        │ 0.45              │ m²                │
 │ Drum Rotation Speed               │ 0.5 – 2.0         │ RPM               │
 │ Operating Vacuum ( $\Delta P$ )     │ 0.6               │ bar (vacuum)      │
 │ Wash Solvent Ratio                │ 2.0               │ L wash / kg cake  │
 └───────────────────────────────────┴───────────────────┴───────────────────┘

# 6. Unit Operation 5: Continuous Drying & Milling Integration

Solid API exiting continuous filtration retains $10–25\%$ residual solvent. Continuous drying is performed using an Active Vibration Paddle Dryer or a Tubular Oscillatory Baffled Dryer.

# 6.1 Heat & Mass Transfer Mechanics in Continuous Drying

Drying occurs in two distinct kinetic regimes:

Constant Rate Period: Removal of unbound surface solvent. Mass transfer rate is governed by gas phase convective transport:

N_A = k_y (y_s - y_b) = \frac{h_c (T_b - T_s)}{\Delta H_{vap}}

Falling Rate Period: Diffusion of bound solvent through the crystalline matrix, governed by Fick’s second law:

\frac{\partial X}{\partial t} = D_{\text{eff}} \frac{\partial^2 X}{\partial z^2}

# 7. Quality Control, Control Strategy & ICH Q13 Compliance

# 7.1 Real-Time Release Testing (RTRT) & Sensor Matrix

Under ICH Q13 (Continuous Manufacturing of Drug Substances and Drug Products), end-to-end lines rely on an integrated PAT matrix for real-time quality assurance:

                  CONTINUOUS PAT MONITORING MATRIX (ICH Q13)
 ┌──────────────────────┬────────────────────────┬────────────────────────┐
 │ Unit Operation       │ PAT Instrument         │ Monitored Quality Attribute│
 ├──────────────────────┼────────────────────────┼────────────────────────┤
 │ Flow Reaction Skid   │ Diamond ATR-FTIR       │ Conversion & Byproducts│
 │ Workup Separator     │ Inline Conductivity    │ Phase purity / Leakage │
 │ MSMPR Crystallizer   │ FBRM & PVM Probe       │ Cord length & Particle PSD│
 │ Continuous Filter    │ NIR Spectroscopy       │ Wash solvent residual  │
 │ Continuous Dryer     │ Gas Phase Mass Spec    │ Loss on Drying (LOD)   │
 └──────────────────────┴────────────────────────┴────────────────────────┘

# 7.2 Out-Of-Spec (OOS) Dynamic Material Diversion Logic

If a PAT sensor detects a process disturbance (e.g., unreacted starting material $> 0.5\%$ or LOD $> 1.0\%$ ), the automated control system activates a 3-way diverter valve within $100\,\text{ms}$ , routing the non-compliant stream to a waste/rework tank.

             DYNAMIC DIVERTER VALVE REJECTION FLOW CHART
  PAT Sensor Reading ──► [ Is CQA in Specification? ]
                                │
                 ┌──────────────┴──────────────┐
                 │ YES                         │ NO
                 ▼                             ▼
        [ Product Stream ]            [ 3-Way Diverter Valve ]
     (Proceeds to Next Unit)          (Triggers Rejection to Rework Tank)

# 8. Summary & Economic Benefits

               ECONOMIC COMPARISON: BATCH VS. END-TO-END CONTINUOUS
 ┌───────────────────────────────────┬───────────────────┬───────────────────┐
 │ Metric                            │ 200 KL Batch Plant│ Integrated E2E-CM │
 ├───────────────────────────────────┼───────────────────┼───────────────────┤
 │ Total Plant Footprint             │  $3,500\,\text{m}^2$ │  $850\,\text{m}^2$  │
 │ Manufacturing Lead Time           │ 45 days           │ 14 hours          │
 │ Work-in-Progress (WIP) Inventory  │  $4.5M │$ 0.2M             │
 │ Solvent Consumption               │  $12.5\,\text{kg/kg}$ │  $3.2\,\text{kg/kg}$ │
 │ Energy Intensity                  │  $45\,\text{kWh/kg}$ │  $12\,\text{kWh/kg}$ │
 │ Quality Rejection Rate            │  $2.5\%$            │  $<0.01\%$          │
 └───────────────────────────────────┴───────────────────┴───────────────────┘

End-to-End Continuous Manufacturing converts API manufacturing from an inefficient, high-inventory batch process into an agile, highly automated, and inherently safe continuous operation.

End-to-End Continuous Manufacturing (E2E-CM) in API Plants: Integrated Flow Synthesis, Crystallization, Filtration & Drying

Peer-Reviewed & PE Verified