Automatic Synthesis of Multimodal Polymers

Abstract

An automatic molecular weight controller is used in semi‐batch reactions to produce final polymers that contain distinct subpopulations with different molecular weight distributions (MWD). While blending polymers with different MWD is frequently used to make multimodal polymer products, a method is introduced here allowing the multimodal product to be made in successive, automatically controlled stages in the same reactor. The method is demonstrated using free radical polymerization of acrylamide to produce widely separated multimodal MWD. Automatic Continuous Online Monitoring of Polymerization reactions with a Control Interface (ACOMP/CI) was used. Weight average molecular weight (Mw) in the ACOMP/CI was continuously measured with multi‐angle light scattering and ultra‐violet absorption. The controller used two principles: both polymerization rate and instantaneous molecular weight are proportional to monomer concentration. By controlling monomer flow rate into the reactor, the controller produced a targeted amount of polymer with a desired first Mw and then automatically introduced chain transfer agent (CTA) into the reactor to produce a second population of polymers with much lower Mw. Subsequently, reactions were performed to produce trimodal populations. This approach is kindred to multi‐stage synthesis of polymers where distinctly different processes are carried out in succeeding stages to produce polymers with highly specific properties.

Future manufacturing and remanufacturing of polymeric materials

“…Servo control, while applicable to continuous polymerization (moving the process from one condition to another during a grade change), but is most applicable to batch and semibatch polymerizations. Here the servo task might be as simple as moving the reactor temperature along a predetermined trajectory, or as complex as bringing a semibatch reactor from initiation to a prespecified target of monomer conversion, average molecular weight and copolymer composition in an optimal way …”

Tuning the molecular weight distribution from atom transfer radical polymerization using deep reinforcement learning

“…This controller is based on online molar mass monitoring techniques (Florenzano et al, 1998) and is able to follow a specific chain growth trajectory with respect to time by controlling the monomer flow rate in a continuous flow reactor. Similar online monitoring techniques have recently enabled controlling the modality of free radical polymerization products, (Leonardi et al, 2017) providing optimal feedback control to acrylamide-water-potassium persulfate polymerization reactors, (Ghadipasha et al, 2017) and monitoring multiple ionic strengths during the synthesis of copolymeric polyelectrolytes. (Wu et al, 2017) However, none of these works attempted to control the precise shape of polymer MWD shapes, nor did they use an artificial intelligence (AI) driven approach to design new materials.…”

“…Each ending MWD is an average over 100 episodes, generated using the trained 1D-CNN controller. Note that this MWD averaging is equivalent to blending polymer products generated in different reactions, (Leonardi et al, 2017) a common practice in both laboratory and industrial polymerization. (Jovanović et al, 2004;Lenzi et al, 2005;DesLauriers et al, 2005;Zhang & Ray, 2002) The trained 1D-CNN agent used in these test runs is that which gave the best performance in the training process, i.e., the neural network weights are those that generated the highest reward during the training process.…”

Coupling of NMR to ACOMP for Terpolymerization Monitoring and Control

“…Recently, this group achieved automatic control of molecular weight in free radical polymerization, automatic production of multi‐modal molecular weight distributions, and then simultaneous control of molecular weight and composition for free radical copolymerization . The latter was achieved for acrylamide (Am) and styrene sulfonate (SS) via continuous monitoring of their individual concentrations using an online UV spectrometer.…”

“…The controllers developed in refs. were essentially model‐free and involved only two constantly measurable proportionality constants for homopolymerizaton control and three constantly measurable proportionality constants for copolymerization control. Aside from these, a formal, model‐based nonlinear controller for homopolyerization molecular weight was also developed and demonstrated …”