The use of model cellulose gel beads to clarify flame-retardant characteristics of layer-by-layer nanocoatings

Layer-by-Layer (LbL) assembled nanocoatings are exploited to impart flame-retardant properties to cellulosic substrates. A model cellulose material can make it possible to investigate an optimal bilayer (BL) range for the deposition of coating while elucidating the main flame-retardant action thus allowing for an efficient design of optimized LbL formulations.

Available online 2 December 2020

( http://creativecommons.org/licenses/by-nc-nd/4.0/ ).

The use of model cellulose gel beads to clarify flame-retardant

characteristics of layer-by-layer nanocoatings

Oruç K¨oklükayaa,* , Rose-Marie Pernilla Karlssona,c, Federico Carosiob, Lars Wågberga,c,*

aDepartment of Fiber and Polymer Technology, KTH Royal Institute of Technology, Teknikringen 56, SE-100 44 Stockholm, Sweden

bDipartimento di Scienza Applicata e Tecnologia, Politecnico di Torino, Alessandria Site Viale Teresa Michel 5, 15121, Alessandria, Italy

cWallenberg Wood Science Center, KTH Royal Institute of Technology, SE-100 44 Stockholm, Sweden

A R T I C L E I N F O

Keywords:

Layer-by-Layer assembly

Flame-retardant

Thermal stability

Cellulose gel beads

A B S T R A C T Layer-by-Layer (LbL) assembled nanocoatings are exploited to impart flame-retardant properties to cellulosic substrates A model cellulose material can make it possible to investigate an optimal bilayer (BL) range for the deposition of coating while elucidating the main flame-retardant action thus allowing for an efficient design of optimized LbL formulations Model cellulose gel beads were prepared by dissolving cellulose-rich fibers followed

by precipitation The beads were LbL-treated with chitosan (CH) and sodium hexametaphosphate (SHMP) The char forming properties were studied using thermal gravimetric analysis The coating increased the char yield in nitrogen to up to 29 % and showed a distinct pattern of micro intumescent behavior upon heating An optimal range of 10–20 BL is observed The well-defined model cellulose gel beads hence introduce a new scientific route both to clarify the fundamental effects of different film components and to optimize the composition of the films

1 Introduction

Cellulose is the most abundant biopolymer on earth (Klemm et al.,

2005), the most common sources of cellulose being wood and cotton

Cotton fibers have been one of the major constituents in textiles, and

wood fibers have a broad application in the pulp and paper industry

Cellulose-based materials are inexpensive, biodegradable and

recy-clable, but the inherent flammable character of cellulose limits its

application or requires flame-retardant treatment for specific

applica-tions Recent developments have shown that treatment of the fiber

surfaces with thin layers of polymers and nanoparticles, through the LbL

technique, can impart excellent flame protection both for textiles (Li,

Schulz, & Grunlan, 2009) and for wood fibers (Koklukaya et al., 2015)

The surfaces of fibers from both cotton and wood are however rough and

chemically heterogeneous and they are not suitable for fundamental

investigations of the assembly of multilayers and the effects of LbL

coatings Different model cellulose surfaces with different degrees of

crystallinity have therefore been developed (Aulin et al., 2009) The

most commonly used films have been prepared by spin coating of

dis-solved cellulose onto smooth silica surfaces (Aulin et al., 2009;

K¨oklükaya et al., 2018) It is not possible to dissolve cellulose in

con-ventional solvents due to its relatively high molecular mass and close

packing of the glucan macromolecule in a crystalline structure How-ever, regenerated cellulose materials can be prepared using solvents

such as N-Methylmorpholine-N-Oxide (NMMO) (Johnson, 1969), cupriethylenediamine (CED) (Schweizer, 1857) or lithium chloride in N, N-dimethylacetamide (LiCl-DMAc) (McCormick, 1981) Through the regeneration of cellulose in suitable liquids, materials with different shapes can be prepared such as films (Wendler et al., 2012), fibers (Woodings, 2003), hydrogels (S Wang et al., 2016), spheres (Oliveira & Glasser, 1996) etc., and the degree of crystallinity of these materials is dependent on the choice of solvent Carrick et al (Carrick et al., 2014) and Karlsson et al (Karlsson et al., 2018) demonstrated the use of LiCl-DMAc to prepare nm smooth cellulose spheres with a crystallinity below 1% Regenerated cellulose fibers widely used in textiles (i.e., lyocell, viscose, and rayon) are prepared using different regeneration processes (Wendler et al., 2012), but the use of native and regenerated cellulose can be limited by the thermal instability and flammability of the cellulose Flame-retardancy can be imparted to cellulosic materials via chemical additives in the wet state (Hall et al., 1999), pad-cure coating (Horrocks, 2011), spray coating (Helmstetter, 1998) and recently, the LbL technique (Holder et al., 2017) The LbL assembly technique has been employed to apply an efficient flame-retardant coating on substrates such as textile (Li, Schulz, & Grunlan, 2009;

* Corresponding authors at: Department of Fiber and Polymer Technology, KTH Royal Institute of Technology, Teknikringen 56, SE-100 44 Stockholm, Sweden

Contents lists available at ScienceDirect Carbohydrate Polymers journal homepage: www.elsevier.com/locate/carbpol

https://doi.org/10.1016/j.carbpol.2020.117468

Received 4 September 2020; Received in revised form 26 November 2020; Accepted 27 November 2020

Srikulkit et al., 2006), polyurethane foam (Kim et al., 2011), wood fibers

(Koklukaya et al., 2015; K¨oklükaya et al., 2020; K¨oklükaya et al., 2018),

and cellulose nanofibril based aerogels (Koklukaya et al., 2017) LbL

treatment is based on the consecutive adsorption of layer constituents at

the solid-liquid interface (Decher, 1997) Different systems have been

investigated to impart micro-intumescent flame-retardant nanocoatings

on cotton fabrics using CH as a carbon source/blowing agent and SHMP

as an acid source (Guin et al., 2014; Leistner et al., 2015; Mateos et al.,

2014) However, there is limited fundamental understanding of the

flame-retardant mechanism and the effects of the LbL-assembled

mul-tilayers prior to their deposition as coatings on wood fiber/cellulose

surfaces The conventional approach is to perform the depositions on the

selected substrate with variable parameters, perform the complete

characterization and then trace back to optimal layer number and

coating mode of action In this way, the mechanisms behind the

improved flame-retardant characteristics of LbL films composed of poly

(allylamine hydrochloride) (PAH) and montmorillonite clay (MMT) on

polyamide 6 as well as for LbL systems comprising of CH and ammonium

polyphosphates (APP) applied to a cotton substrate could be indirectly

identified (Apaydin et al., 2014; Jimenez et al., 2016) It was shown that

the flame-retardant mode of action of (PAH/MMT) coating occurred in

the condensed phase The 40 BLs of (PAH/MMT) coating protected the

underlying polyamide 6 from an external heat flux of 25 kW/m2

(Apaydin et al., 2014) It was also shown for a (CH/APP) multilayer

coating that the flame-retardant behavior was due to a combination of a

condensed phase forming an aromatic char layer and a gas phase

releasing non-flammable volatiles that promote the micro-intumescence

phenomenon (Jimenez et al., 2016) Based on earlier investigations it is

apparent that the common approach is to investigate the optimal

deposition conditions and flame-retardant mechanism after a complete

characterization of the treated substrates (Apaydin et al., 2014; Guin

et al., 2014; Jimenez et al., 2016; Mateos et al., 2014) Within this

context, model substrates such as silicon oxide have also been employed

in order to investigate the compositional and morphological changes

occurring within the coating after the exposure to a flame or to a heat

flux (Koklukaya et al., 2017; Maddalena et al., 2018) This approach has

the limitation of focusing only on the coating disregarding the effects of

the substrate that is replaced by silicon oxide Thus, although the earlier

studies helped to identify the overall flame-retardant effect of LbL

multilayers, the use of a small scale and a controlled preliminary

screening approach involving the substrate of interest would have

allowed for the optimal design of the coating architecture and

compo-sition while providing a meaningful insight on the crucial interactions

occurring during combustion between the deposited LbL coating and the

substrate In order to address such questions, we propose a simple and

yet effective strategy for the study and design of a flame-retardant LbL

assembly of nanocoatings directly on model cellulose beads To this aim,

we have used dissolved carboxymethylated fibers to prepare

cellulose-based hydrogel beads to be used as model cellulose substrates

in combination with the LbL technique to deposit intumescent coatings

of CH and SHMP This system is a good candidate for studying the

fundamental processes behind the development of intumescence coating

as it has already been used to confer flame-retardant properties to cotton

(Guin et al., 2014) These authors reported an improved cellulose char

formation combined with the formation of a barrier consisting of

sub-micron-sized bubbles as the main mechanism for optimal

flame-retardancy (Guin et al., 2014; Jimenez et al., 2016) In the present

work we are using our model system to investigate a much deeper

un-derstanding of the fundamentals behind the flame-retardant action of

the CH/SHMP system Model experiments were also performed using

silicon oxide model surfaces (Carosio et al., 2018; Koklukaya et al.,

changes of the nanocoating during degradation pointing out a micro-intumescent behavior The results also demonstrate the excellent applicability of the cellulose beads as model substrates for cellulose rich materials in a variety of fundamental studies

2 Experimental

2.1 Materials

The cellulose fibers employed in this study were obtained from a dissolving grade pulp supplied by Domsj¨o Fabriker AB, ¨Ornsk¨oldsvik, Sweden The cellulose content of the pulp was 93 % and the degree of

polymerization was about 780 (provided by the manufacturer) N,N- dimethylacetamide (DMAc) (>99.5 %, GC grade), lithium chloride (LiCl), and acetic acid (Sigma-Aldrich) were used as received CH (Mw =

60 000, 95 % deacetylation) was supplied by GTC Union Corp., Qingdao, China, and SHMP (crystalline, +200 mesh, 96 %) was obtained from Sigma-Aldrich, Stockholm, Sweden Poly(vinyl amine) (PVAm), com-mercial name Xelorex 6300, was supplied by BASF PVAm was dialyzed and freeze-dried prior to use Monochloroacetic acid, methanol, iso-propanol, ethanol, HCl, NaOH, and NaCl were all analytical grade pur-chased from Merck, Stockholm Sweden

2.2 Cellulose gel bead preparation

The cellulose fibers were first carboxymethylated following to the method previously described by Wågberg et al (Wågberg et al., 2008) and the degree of substitution (D.S) was calculated by conductometric titration (Katz & Beatson, 1984) to be 0.13 which corresponds to a charge density of 795 μeqv/g 1 g of the carboxymethylated pulp was then dissolved in 100 mL solution of 7 wt% LiCl/DMAc following the steps previously described by Karlsson et al (Karlsson et al., 2018) and Carrick et al (Carrick et al., 2014) The water in the pulp was first sol-vent exchanged by displacement with ethanol and the ethanol was subsequently exchanged with DMAc using a filtration procedure Each solvent was displaced over a period of two days during which the solvent was changed at least twice per day After this first step, the DMAc in which the pulp was to be dissolved was dehydrated by heating and keeping it for 30 min at a temperature of 105 ◦C The LiCl was also dehydrated during this 30 min in an oven at 105 ◦C After the dehy-dration, the DMAc was allowed to cool and the LiCl was added and dissolved The pulp was added to the DMAc/LiCl solution at a temper-ature of ca 40 ◦C and then instantly placed in a 5 ◦C fridge and stirred with a magnetic stirrer overnight After about 24 h, the solution was clear The solution was then filtered using a 0.45 μm PTFE syringe filter and the filtrate was employed to form gel beads by drop-wise precipi-tation through a needle of 0.64 mm into about 95 mL of a non-solvent consisting of 80 mL 0.03 M HCl (aq) with 15 mL ethanol The gel beads formed were allowed to rest in the bottom of the beaker at 5 ◦C for

24 h The non-solvent was then replaced with deionized water by decanting about 80 mL of the non-solvent four times during two days and stepwise decreasing the concentration of HCl The gel beads were then washed with deionized water for one week in order to remove any residual DMAc/LiCl The wet cellulose gel beads have an average diameter of 2.7 mm and after drying the average diameter of beads was 0.6 mm Prior to LbL treatment, the gel beads were dried at 23 ◦C and 50

% RH

2.3 Model LiCl/DMAc cellulose films

Cellulose dissolved in a solution of DMAc/LiCl was used to prepare

The silicon wafers and QCM-D crystals were cleaned by rinsing them in a

sequence of Milli-Q water, ethanol and Milli-Q water and then drying

them with nitrogen gas The surfaces were plasma treated for 3 min in a

plasma oven (PDC-002, Harrick Scientific Inc.) at 30 W under reduced

air pressure Prior to spin coating of the cellulose solution, silicon wafers

were treated with 0.1 g/L PVAm solution at pH 7.5 for 15 min to form an

anchoring layer for the cellulose and then rinsed with Milli-Q water and

dried with a stream of nitrogen Spin-coating was performed using a spin

coater (KW-4A-2, Chemat Technology, Northridge, CA, USA) and the

cellulose solution was placed onto a PVAm treated silicon surface on the

spin-coater disk Spin coating was performed at 3 000 rpm for 30 s

These model cellulose surfaces were precipitated by immersion in

Milli-Q water The substrates were then placed in Milli-Q water to

remove excess solvent and LiCl Finally, the substrates were dried with

nitrogen and stored in a desiccator until further use The dry thickness of

the non-crystalline cellulose film was measured by AFM to be 38 ± 2 nm

2.4 Layer-by-Layer deposition

CH solution (1 g/L) was prepared in 1 v/v% acetic acid, and a SHMP

solution (5 g/L) was prepared in Milli-Q water (18.2 MΩ cm Milli-Q

grade water Synergy 185, Millipore Bellerica, USA) Both solutions

were stirred with a magnetic stirrer for 24 h to ensure complete

disso-lution and the pH of the sodisso-lutions was then adjusted to pH 5 using 5 M

NaOH for CH and 1 M HCl for SHMP and the electrolyte concentration

was adjusted to 10 mM NaCl The silicon wafers were cleaned according

to the method previously described with Milli-Q water, ethanol and

Milli-Q water and dried with nitrogen gas (Aulin et al., 2008) The

sil-icon wafers were then placed in an air plasma cleaner (PDS 002, Harrick

Scientific Corp.) for 3 min in order to clean and activate the surface prior

to LbL deposition The silicon wafers or model cellulose surfaces were

alternately dipped into polyelectrolyte solutions in the order CH and

SHMP using an automatic dipping robot (StratoSequence VI, nanoStrata

Inc., Tallahassee, Florida, USA) The adsorption time for the first bilayer

was 5 min in each solution in order to achieve a uniform deposition,

while the time for the rest of the depositions was 1 min The substrates

were rinsed with Milli-Q water (pH 5) three times between each

depo-sition for 1 min in each rinsing step without intermediate drying For LbL

deposition on cellulose gel beads, a home-made filtration set up was

used (Fig 1) Prior to LbL deposition, the cellulose beads were washed

according to the previously described procedure, the carboxyl groups

were converted to their sodium form (Wågberg & Bj¨orklund, 1993), and

the beads were dried at 23 ◦C and 50 % RH Dried cellulose beads were

immersed in cationic CH solution for 5 min, after which the solution was

filtered by suction and the beads were rinsed twice with Milli-Q water

(pH 5) to ensure removal of loosely adhered and excess polymer The

beads were then exposed to anionic SHMP solution by filling the

container with a solution and allowing adsorption for 5 min The

solu-tion was then filtered by applying vacuum pressure and the beads were

rinsed twice with Milli-Q water (pH 5) One such sequence of deposition

represents one bilayer (BL) This process was repeated with 1 min adsorption times until the desired number of bilayers had been depos-ited Containers and filters were replaced after deposition of every 20 BL

to avoid any complex formation Coated beads were placed on a Teflon surface after LbL-treatment and dried at 23 ◦C and 50 % RH Photograph

of cellulose gel bead before and after the LbL-treatment is shown in supporting information Figure S1

2.5 Thin film characterization 2.5.1 Quartz crystal microbalance with dissipation

A Quartz Crystal Microbalance with Dissipation (QCM-D, Q-Sense

AB, G¨oteborg, Sweden) was used to estimate both the amount of adsorbed polyelectrolyte with associated water and the viscoelastic properties of the adsorbed film The normalized frequency change can

be related to the adsorbed mass of polyelectrolyte and water and the energy dissipation can be related to the viscoelastic properties of the film (Rodahl et al., 1995) The adsorption of polyelectrolytes and the rinsing steps were monitored until saturation was reached

2.5.2 Atomic force microscopy

An Atomic force microscope (AFM), Nanoscope IIIa (Bruker AXS, Santa Barbara, CA) was used to investigate the surface topography, roughness, and thickness of the multilayer films deposited on model cellulose surfaces prepared on silicon wafers The films were scratched with a sharp blade in the dry state The thickness was measured before and after LbL treatment to determine the thickness of the films formed E and J-type piezoelectric scanners and Scanasyst cantilevers with a nominal resonance frequency of 70 kHz and a 0.4 N/m spring constant were used to scan the surfaces in air The surface roughness value was calculated from acquired images with an area of 2 × 2 μ m2

2.5.3 Nitrogen analysis

The ANTEK 7000 nitrogen analyzer (Antek Instruments, Houston,

TX, USA) was used to measure the nitrogen content of chitosan adsorbed

on the cellulose beads The method is based on combustion of the sample (c.a 5 mg) at 1050 ◦C in an oxygen-poor atmosphere where nitrogen is oxidized to NO before being further oxidized to excited NO2 in ozone The light emitted when the excited NO2 is converted to its standard state

is detected by a photomultiplier tube The system is calibrated with a known amount of CH (Supporting information Figure S2)

2.5.4 Fourier transform infrared spectrometry

Spectra of cellulose beads were obtained using a Perkin-Elmer Spectrum 2000 FTIR with an attenuated total reflectance crystal acces-sory (Golden Gate) ATR-FTIR spectra were recorded in the 4000− 600

cm− 1 region at a resolution of 4.0 cm− 1 and using 16 scans

2.5.5 Thermogravimetric analysis

The thermal degradation of the untreated and LbL-treated cellulose

Fig 1 a) Photograph of cellulose gel bead in wet swollen state (on

the right) and once dried then again swollen in water (on the left), b) Schematic description of the LbL assembly on cellulose gel beads Beads were treated with cationic chitosan (CH) and anionic sodium hexametaphosphate (SHMP) The process was repeated in order to deposit 100 BL The polyelectrolyte concentration was 1 g/L for CH and 5 g/L for SHMP, both in a 10 mM NaCl aqueous solution at pH 5 The rinsing solution was Milli-Q water at pH 5 and c) Schematic of cellulose bead before and after LbL assembly

beads was investigated by thermogravimetric analysis (TGA) (Mettler

Toledo TGA/DSC, Stockholm, Sweden) The samples (5 ± 1 mg) were

placed in 70 μL aluminium oxide crucibles and heated at a rate of 10 ◦C/

min from 40 to 800 ◦C in nitrogen at a flow rate of 50 mL/min

2.5.6 Heating element

The thermal degradation behavior of untreated and LbL treated

cellulose beads were monitored using a high-speed camera (IDT N4M-

S3) with a 2× magnification microscope lens The samples were

placed in a small ceramic crucible and covered with a microscope slide

cover slip A light emitting diode light source (IDT 7 LED) was used to

illuminate the samples The ceramic crucible was placed on a flat

heating element (d 10.8 × 2 mm/24 V/50 W/750 ◦C/Button heater,

Rauschert Steinbach GmbH, Germany) with heating rate of 300 ◦C/min

when the temperature was set to ~370 ◦C The changes in the structure

of the samples were recorded using high-speed camera at a frame rate of

100 frames per second A schematic of the experimental setup is shown

in Figure S3

2.5.7 Scanning Electron microscopy and energy dispersive X-ray analysis

A field emission scanning electron microscope (FE-SEM, Hitachi S-

4800) was used to investigate the surface morphology of cellulose beads

before and after the LbL treatment The residues from the heating

element test were also investigated with FE-SEM to study the change in

morphology Test pieces were coated with a 5 nm thick platinum/

palladium layer using a Cressington 208 HR high-resolution sputter

coater The presence of phosphorus in the LbL-treated cellulose beads

before and after the heat element test and TGA was performed using an

Inca (Oxford Instruments, X-MAX N80) energy dispersive X-ray

spec-trometer (EDX) Dried cellulose beads were mounted on the specimen

holder and then cut in half using a razor blade Samples for EDX were not

sputter coated with Pt/Pd

3 Results and discussion

3.1 Monitoring the build-up of multilayer films on flat model surfaces and on cellulose beads

The LbL formation of multilayer films consisting of CH and SHMP on model cellulose surface was investigated using QCM-D Fig 2 shows the results from the QCM-D measurements, where the normalized frequency shift for the third overtone and the change in the dissipation are shown

as functions of the number of adsorbed layers

The frequency shift during the adsorption of CH showed an increase followed by a decrease during SHMP adsorption, as shown in Fig 2a The energy dissipation data (ΔD) showed a significant decrease during adsorption of the CH layer followed by an increase during the adsorption

of the SHMP layer Benselfelt et al reported a similar behavior for the adsorption of a multilayer film of PDADMAC/PSS on a model cellulose surface (Benselfelt et al., 2017) Earlier studies have also clearly shown a deswelling of cellulose film due to polyelectrolyte adsorption (Benselfelt

et al., 2017; Enarsson & Wågberg, 2008; Notley, 2008; Wang et al.,

2011; Vuoriluoto et al., 2015) It can therefore be suggested that the detected changes in the QCM-D measurements following the adsorption

of CH are due to a deswelling of the highly charged cellulose film on the QCM crystal due to a neutralization of the charges of the cellulose by the adsorbed CH similar to earlier results (Benselfelt et al., 2017) The decrease in the pH accompanying the addition of CH will naturally add

to this effect but, as noted earlier (Xie & Granick, 2002), the charges of the CH will also increase the degree of dissociation of the carboxyl groups in the cellulose, which means that the effect of the pH will probably not be the dominating cause of the deswelling When the SHMP

was added, there was a significant decrease in frequency and a

concomitant increase in the dissipation, which indicate that when the

SHMP is adsorbed, the charge of the chitosan is efficiently compensated

and probably over-compensated, which means that the cellulose film is

again able to swell, and this swelling is much more significant compared

to earlier results (Benselfelt et al., 2017) The trends for the subsequent

layers are the same, showing both a steady build-up of the LbL film and a

reversible swelling and deswelling of the cellulose film which is a clear

demonstration of the dynamics of the LbL assembly during the build of

the films which indeed is important for the development of the

prop-erties of the films

To add further information to the QCM-D measurements and to

quantify the amount of polymers in the adsorbed layers the cellulose

beads were also used as a substrate for LbL deposition and the amount of

CH adsorbed was determined using nitrogen analysis The results

showed a steady increase in adsorbed amount during the build-up of 100

BL supporting the results of the QCM-D measurements regarding the

steady build-up of LbLs on the cellulose surface (Fig 3) The initial value

for the reference sample is probably due to residual nitrogen containing

cellulose solvent in the prepared beads and this value could be

sub-tracted from the other layers since the amount of adsorbed polymer was

determined from a calibration curve using the CH (Figure S2)

The chemical composition of the LbL coatings deposited on cellulose

beads has been assessed qualitatively using FTIR spectroscopy in an

attenuated total reflection (ATR) configuration Fig 4 shows the spectra

of cellulose beads, CH, SHMP and LbL-treated cellulose beads The

characteristic peaks of cellulose are described in supporting

informa-tion The peak observed at 1734 cm− 1 was ascribed to protonated

car-boxylic acids The LbL-treated cellulose beads showed peaks at 1655

cm− 1 ascribed to the carbonyl (C=O), at 1592 cm− 1 ascribed to NH2, and

at 1153 cm− 1, 1063 cm− 1, and 1028 cm− 1 ascribed to stretching

vi-brations of C–O–C in glucosidic bonds of CH (Osman & Arof, 2003) The

presence of SHMP in the coating was shown by two strong signals at

1250 cm− 1 and 865 cm− 1 corresponding to stretching of P=O and P–O–P

groups (Drevelle et al., 2005) The absorbance intensity of two peaks

increased as the number of bilayers increased indicating the build-up of

the multilayer Further evidence of the thin film growth was provided by

the fact that the weak signal at 1734 cm− 1, related to C=O stretching

vibrations in the carboxylic group present on the beads slowly

dis-appeared during the LbL deposition The absorbance peaks of LbL thin

film starts to dominate over the absorbance peaks of cellulose already at

10 BL of deposition

3.2 Thickness and roughness of the films on model surfaces

Model cellulose surfaces prepared on silicon wafers were used as substrates for multilayer film formation and the LbL-assembled films were imaged using AFM to characterize the morphology, roughness, and thickness of the dry films The height images of the CH/SHMP films are shown in Fig 5 The thickness of the films was measured using AFM by scanning the film scratched with a scalpel (Figure S4)

The thickness and roughness are shown in Fig 6 as functions of the number of deposited bilayers

The films deposited on the model cellulose surfaces were somewhat thicker than those deposited on the silicon wafers and the LbL film build-

up showed a linear increase in the thickness up to 20 BL, after which the increase in thickness was non-linear with additional BL deposition It has been suggested that the change in the LbL growth with the number of BLs can be due to an in and out diffusion of polyelectrolytes in the LbLs (Guin et al., 2014; Picart et al., 2001) or a type of island growth with increasing number of BLs (Haynie et al., 2011) where an initial un-evenness propagates and small islands grow into larger islands as the number of BL increases After passing 20 BL, there is a super-linear growth (Abdelkebir et al., 2011) and the roughness was larger for 10 and 20 BL of films deposited on silicon wafers than those deposited on model cellulose surfaces, but the roughness was similar for 50 and 100

BL of film on both surfaces and the formed layers were indeed very smooth A more detailed analysis of the surfaces also shows that there is

a granular morphology of 10 and 20 BL of CH/SHMP film on the silicon oxide surfaces, as shown in Figure S5, and also on the cellulose model surfaces as shown in Fig 5 It can also be seen that the smaller granular shape of the surfaces changes into a larger scale unevenness at around 50

BL, which results in a lower roughness value but also fits with the island growth model (Haynie et al., 2011) This behavior has already been observed for non-linear LbL systems and has been ascribed to the in and out diffusion of polyelectrolytes in the LbL structure (Picart et al., 2002)

It is not possible to establish the molecular reason for the change in growth detected when passing 20 BL for the present system, but it is clear that there is a steady growth of the LbLs with the number of deposition steps and that the granular structure of the surfaces changes

to a more even surface leaving a rather flat and flaw-free surface which

is probably essential for good flame-retardancy of the treated surfaces

3.3 Thermogravimetric analysis of cellulose beads

Thermogravimetric analysis in nitrogen was used in order to eval-uate the effect of the LbL coating on the char forming ability of the cellulose gel beads in an oxygen depleted environment This approach can provide preliminary hints on the pyrolysis occurring in the condensed phase of a burning material where the presence of a protec-tive coating results in an essentially anaerobic atmosphere and reduced heating rates Fig 7 shows the weight loss (TG) and the derivative of the weight loss (dTG) as a function of temperature and Table 1 presents the degradation temperatures and residual amounts of reference and LbL- treated beads

The untreated and LbL-treated cellulose beads show similar thermal degradation processes The initial weight loss observed at 100 ◦C is attributed to the dehydration of water adsorbed by the coating The first significant loss of mass observed at 246 ◦C is attributed to dehydration due to water entrapped within the cellulose gel beads and depolymer-ization of non-crystalline cellulose leading to an aliphatic char The rate

of mass loss was lower in the LbL-treated than in the reference beads It

is suggested that this is due to the charring properties of the LbL coating since the rate at which thermal energy reaches the surface of cellulose beads is reduced by the LbL film (Hribernik et al., 2007) A second degradation step occurs at 308 ◦C The LbL-treated cellulose beads exhibit no early degradation, which is generally reported to be due to the presence of phosphorus (Guin et al., 2014) A possible explanation is that the temperature at which the phosphorus compound catalyzes the

Fig 3 Total amount of CH adsorbed on cellulose beads as a function of number

of bilayers deposited, determined by nitrogen analysis using a calibration curve

for the CH The polyelectrolyte concentrations were 1 g/L for CH and 5 g/L for

SHMP, both in a 10 mM NaCl aqueous solution at pH 5 The rinsing solution

was Milli-Q water at pH 5

Fig 4 FTIR spectra of a) cellulose gel beads, b) CH, c) SHMP and d) CH/SHMP treated cellulose beads (10, 20, 50 and 100 BL) The polyelectrolyte concentrations

were 1 g/L for CH and 5 g/L for SHMP, both in a 10 mM NaCl aqueous solution at pH 5 The rinsing solution was Milli-Q water at pH 5

dehydration of cellulose overlaps the degradation temperature of

non-crystalline cellulose The amount of residue found at 800 ◦C

increased as the number of bilayers deposited increased This behavior

can be attributed to the presence of phosphate groups in the SHMP

which favor the dehydration of CH towards the formation of an aromatic

char which acts as a thermal barrier to limit mass and heat transfer to the

cellulose beads (Carosio et al., 2015) In addition, an optimum in char

forming efficiency is clearly observable in the 10–20 BL range as the

increase of the deposited BL to 50 and 100 yields diminish returns in

terms of final residues (Table 1) This suggests that a flame-retardant

application of this system should target a BL number in between 10

and 20 BL and is in good agreement with a previous study where 17 BL

deposited on cotton yielded optimal flame-retardant properties (Guin

et al., 2014)

In order to mimic the exposure to a heating rate relevant to a fire

scenario, the beads were further characterized at high heating rates

(~300 ◦C/min) by using a specially designed heating element in which the untreated and LbL-treated cellulose beads were heated to a high temperature (~370 ◦C) and monitored with a high speed camera in order to assess the steep degradation curve observed in TGA evaluations (Figure S6) 370 ◦C was selected in order to ensure completion of all the main degradation steps observed by TGA All the samples immediately began to exhibit thermal degradation resulting in the formation of a char layer During the degradation, sudden movements (e.g., jumping of the sample) were observed which were attributed to the release of entrap-ped water vapor within the cellulose beads and in the LbL structures The residues from this test were then investigated using SEM in order to assess the coating morphology Fig 8 shows SEM images of untreated and LbL-treated beads before and after heat application

The untreated beads had a relatively smooth morphology After deposition of 10 BLs of CH/SHMP, the beads appeared to have a wrin-kled morphology A possible explanation of this structural change is the difference in modulus of cellulose and of the coating (Nolte et al., 2005; Stafford et al., 2004) More specifically, the cellulose beads were dry prior to the LbL treatment but during the treatment process, they became completely swollen due to the presence of water Having different moduli, the beads and the thin LbL films create a stress which forms the wrinkled morphology (Chan & Crosby, 2011) upon drying as shown in Fig 8 A further increase in BL number results in an increased coating thickness as shown by AFM, which consequently forms larger wrinkles in the coating After the application of heat, both the untreated and LbL-treated beads maintain their shapes The charring layer of the

Fig 6 The average a) thickness and b) roughness values of CH/SHMP films deposited on silicon oxide and model cellulose surfaces as functions of the number of

bilayers deposited The polyelectrolyte concentrations were 1 g/L for CH and 5 g/L for SHMP, both in a 10 mM NaCl aqueous solution at pH 5 The rinsing solution was Milli-Q water at pH 5

Fig 7 a) Weight loss (TG) and b) derivative of weight loss (dTG) for untreated and CH/SHMP treated cellulose beads in a nitrogen atmosphere

Table 1

TGA data for untreated and CH/SHMP treated cellulose beads in a nitrogen

atmosphere

Sample T max1 [ ◦ C] T max2 [ ◦ C] Residue [%]

Cellulose bead 245 308 6

untreated cellulose bead exhibits obvious cracks and voids due to

cel-lulose pyrolysis and the release of volatile compounds On the contrary,

LbL-treated beads show a unique structure which is generally defined as

micro-intumescent bubbling (Carosio et al., 2015; Li et al., 2011) The

formed sub-micronic bubbles become larger and more distinguishable as

the number of BLs deposited is increased The reason for this change is

not exactly known but it can be suggested that the bubbles are formed as

a consequence of the release of volatile gases inside the beads, and that

this in turn creates a stress on the coating layer which yields by creating

bubbles This behavior can be related to the thickness and barrier

properties of the films formed More extensive model experiments are

needed to clarify these mechanisms, but the results show the potential of

using the beads to establish the molecular mechanism for different

LbL-treatments of cellulose surfaces In addition, it is worth highlighting

that a nearly identical post combustion morphology has been observed

on cotton fabrics treated by the same CH/SHMP assembly (Guin et al.,

2014) This further shows the ability of the approach developed in this

work in predicting the coating behavior in real scale testing conditions

Fig 9 shows cross-sections of untreated and 100 BL treated cellulose

beads together with 50 BL treated cellulose beads after heat treatment

protective layer around the bead even after the heat application This is further shown by the elemental analysis line spectrum of the cross- section of a 20 BL treated cellulose bead after thermal gravimetric analysis (Figure S7) Interestingly, no phosphorus was detected in the core of the cellulose bead, indicating that the phosphate action upon heating took place only on the surface of the cellulose bead where the coating was located This suggests that, upon heating, no migration of the phosphate occurs inside the beads and that the main action of SHMP

is to mainly favor the coating char formation with limited effects on cellulose The formation of a protective barrier can then limit heat transfer and promote cellulose charring as it is well known that the char forming ability of cellulose is inversely proportional to the heating rate (Alongi et al., 2013) This further explains the observed diminishing returns in performances upon increasing the number of BL as observed

by TGA In the 10–20 BL range the assembly produces a continuous and thick enough coating capable of providing good thermal shielding per-formances; since there is no phosphate migration, increasing the amount

of SHMP by adding more layers only slightly improves the coating performances

Fig 8 SEM images of untreated and CH/SHMP treated cellulose beads before (the left-hand column) and after (the right-hand column) heat application a)

Un-treated cellulose bead, b) 10 BL, c) 20 BL, d) 50 BL, and e) 100 BL The higher magnification SEM images of indicated area by square frames are shown adjacent to corresponding SEM image of cellulose bead The polyelectrolyte concentrations were 1 g/L for CH and 5 g/L for SHMP, both in a 10 mM NaCl aqueous solution at pH

5 The rinsing solution was Milli-Q water at pH 5

well-defined, non-crystalline cellulose gel beads as a model substrate to

examine the molecular mechanism behind the effect of the multilayer

coating on the thermal degradation of cellulose FTIR measurements

showed that absorbance peaks of pristine cellulose were dominated by

the absorbance peaks of coating after deposition of 10 BL

Thermogra-vimetric analysis revealed that cellulose beads coated with CH/SHMP

films exhibit a degradation behavior different from that of the uncoated

reference beads The multilayer coating of CH/SHMP due to synergetic

effect enhanced the char formation by favoring the dehydration of

lulose and the char formed subsequently protected the underlying

cel-lulose resulting in a residue as high as 29 % at 800 ◦C for 100 BL coated

beads under a nitrogen atmosphere In addition, the amount of residue

significantly increased by a factor of 3.5 after only 10 BLs had been

deposited but a further increase in the BL number did not show a similar

increase SEM images and EDX spectra show the formation of a

micro-intumescent swollen char layer located on the surface of the

LbL-treated beads A correlation of the observed results with previously

reported literature (Apaydin et al., 2014; Holder et al., 2017; Jimenez

et al., 2016) dealing with the use of LbL assembled coatings for

flame-retardancy clearly demonstrates the effectiveness of the proposed

approach in providing meaningful insights on optimal BL range, coating

mechanism and microstructure changes upon heating The proposed

colloidal approach has never been investigated before since common

practice of previously reported literature was to investigate the optimal

deposition conditions and flame-retardant mechanism after a complete

characterization of the treated substrates (Apaydin et al., 2014; Guin

et al., 2014; Jimenez et al., 2016; Mateos et al., 2014) Conversely, this

model material provides an excellent experimental platform for

in-vestigations aimed at a clear understanding of the effect of different

surface treatments on the thermal degradation of cellulose Further

developments of the proposed approach might involve the design of

cellulose beads characterized by tunable degree of crystallinity (H Li

et al., 2020) as well as the study of different LbL assembly encompassing

nanoparticles and the implementation of advanced characterization

techniques aiming at a deeper investigation of the molecular scale

mechanisms of the assembly

Author contributions

The manuscript was written through the contributions of all the

authors All the authors have given their approval to the final version of the manuscript

CRediT authorship contribution statement Oruç K¨oklükaya: Investigation, Writing - original draft Rose-Marie Pernilla Karlsson: Investigation, Writing - review & editing Federico Carosio: Investigation, Validation, Writing - review & editing Lars Wågberg: Supervision, Validation, Writing - review & editing Declaration of Competing Interest

The authors declare no competing financial interest

Acknowledgment

Lars Wågberg, Oruç K¨oklükaya, and Federico Carosio acknowledge financial support from SSF (The Swedish Foundation for Strategic Research) and Lars Wågberg and Rose-Marie Pernilla Karlsson also acknowledge The Wallenberg Wood Science Centre for financial support

Appendix A Supplementary data

Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.carbpol.2020.117468

References

Abdelkebir, K., Gaudi`ere, F., Morin-Grognet, S., Coquerel, G., Labat, B., Atmani, H., & Ladam, G (2011) Evidence of different growth regimes coexisting within

biomimetic layer-by-layer films Soft Matter, 7(19), 9197–9205

Alongi, J., Camino, G., & Malucelli, G (2013) Heating rate effect on char yield from

cotton, poly(ethylene terephthalate) and blend fabrics Carbohydrate Polymers, 92(2),

1327–1334

Apaydin, K., Laachachi, A., Fouquet, T., Jimenez, M., Bourbigot, S., & Ruch, D (2014) Mechanistic investigation of a flame retardant coating made by layer-by-layer

assembly RSC Advances, 4(82), 43326–43334

Aulin, C., Josefsson, P., & Wågberg, L (2009) Nanoscale cellulose films with different crystallinities and mesostructures: Their surface properties and interaction with

water Langmuir, 25(13), 7675–7685

Fig 9 Cross-section SEM images and corresponding EDX elemental mapping of untreated, 100 BL treated and, after heat application, 50 BL treated cellulose beads

Aulin, C., Shchukarev, A., Lindqvist, J., Malmstr¨om, E., Wågberg, L., & Lindstr¨om, T

(2008) Wetting kinetics of oil mixtures on fluorinated model cellulose surfaces

Journal of Colloid and Interface Science, 317(2), 556–567

Benselfelt, T., Pettersson, T., & Wågberg, L (2017) Influence of surface charge density

and morphology on the formation of polyelectrolyte multilayers on smooth charged

cellulose surfaces Langmuir, 33(4), 968–979

Carosio, F., Ghanadpour, M., Alongi, J., & Wågberg, L (2018) Layer-by-layer-assembled

chitosan/phosphorylated cellulose nanofibrils as a bio-based and flame protecting

nano-exoskeleton on PU foams Carbohydrate Polymers, 202, 479–487

Carosio, F., Negrell-Guirao, C., Di Blasio, A., Alongi, J., David, G., & Camino, G (2015)

Tunable thermal and flame response of phosphonated oligoallylamines layer by

layer assemblies on cotton Carbohydrate Polymers, 115, 752–759

Carrick, C., Pendergraph, S A., & Wågberg, L (2014) Nanometer smooth, macroscopic

spherical cellulose probes for contact adhesion measurements ACS Applied Materials

& Interfaces, 6(23), 20928–20935

Chan, E P., & Crosby, A J (2011) Wrinkling polymers for surface structure control and

functionality Polymer thin films (pp 141–161) World Scientific

Decher, G (1997) Fuzzy nanoassemblies: Toward layered polymeric multicomposites

Science, 277(5330), 1232–1237

Drevelle, C., Lefebvre, J., Duquesne, S., Le Bras, M., Poutch, F., Vouters, M., &

Magniez, C (2005) Thermal and fire behaviour of ammonium polyphosphate/

acrylic coated cotton/PESFR fabric Polymer Degradation and Stability, 88(1),

130–137

Enarsson, L.-E., & Wågberg, L (2008) Polyelectrolyte adsorption on thin cellulose films

studied with reflectometry and quartz crystal microgravimetry with dissipation

Biomacromolecules, 10(1), 134–141

Eriksson, J., Malmsten, M., Tiberg, F., Callisen, T H., Damhus, T., & Johansen, K S

(2005) Enzymatic degradation of model cellulose films Journal of Colloid and

Interface Science, 284(1), 99–106

Guin, T., Krecker, M., Milhorn, A., & Grunlan, J C (2014) Maintaining hand and

improving fire resistance of cotton fabric through ultrasonication rinsing of

multilayer nanocoating Cellulose, 21(4), 3023–3030

Hall, M., Horrocks, A., & Seddon, H (1999) The flammability of Lyocell Polymer

Degradation and Stability, 64(3), 505–510

Haynie, D T., Cho, E., & Waduge, P (2011) “In and out diffusion” hypothesis of

exponential multilayer film buildup revisited Langmuir, 27(9), 5700–5704

Helmstetter, J G (1998) Fire resistant coatings for cellulosic materials.

Holder, K M., Smith, R J., & Grunlan, J C (2017) A review of flame retardant

nanocoatings prepared using layer-by-layer assembly of polyelectrolytes Journal of

Materials Science, 1–37

Horrocks, A R (2011) Flame retardant challenges for textiles and fibres: New chemistry

versus innovatory solutions Polymer Degradation and Stability, 96(3), 377–392

Hribernik, S., Smole, M S., Kleinschek, K S., Bele, M., Jamnik, J., & Gaberscek, M

(2007) Flame retardant activity of SiO2-coated regenerated cellulose fibres Polymer

Degradation and Stability, 92(11), 1957–1965

Jimenez, M., Guin, T., Bellayer, S., Dupretz, R., Bourbigot, S., & Grunlan, J C (2016)

Microintumescent mechanism of flame-retardant water-based chitosan–ammonium

polyphosphate multilayer nanocoating on cotton fabric Journal of Applied Polymer

Science, 133(32)

Johnson, D L (1969) Process for strengthening swellable fibrous material with an amine

oxide and the resulting material to Eastman Kodak Co.)

Karlsson, R.-M P., Larsson, P T., Yu, S., Pendergraph, S A., Pettersson, T., Hellwig, J., &

Wågberg, L (2018) Carbohydrate gel beads as model probes for quantifying non-

ionic and ionic contributions behind the swelling of delignified plant fibers Journal

of Colloid and Interface Science

Katz, S., & Beatson, R P (1984) The determination of strong and weak acidic groups in

sulfite pulps Svensk Papperstidning, 87(6), 48–53

Kim, Y S., Davis, R., Cain, A A., & Grunlan, J C (2011) Development of layer-by-layer

assembled carbon nanofiber-filled coatings to reduce polyurethane foam

flammability Polymer, 52(13), 2847–2855

Klemm, D., Heublein, B., Fink, H P., & Bohn, A (2005) Cellulose: Fascinating

biopolymer and sustainable raw material Angewandte Chemie International Edition,

44(22), 3358–3393

Koklukaya, O., Carosio, F., Grunlan, J C., & Wagberg, L (2015) Flame-retardant paper

from wood fibers functionalized via layer-by-layer assembly ACS Applied Materials &

Interfaces, 7(42), 23750–23759

Koklukaya, O., Carosio, F., & Wågberg, L (2017) Superior flame-resistant cellulose

nanofibril aerogels modified with hybrid layer-by-Layer coatings ACS Applied

Materials & Interfaces, 9(34), 29082–29092

K¨oklükaya, O., Carosio, F., Dur´an, V L., & Wågberg, L (2020) Layer-by-layer modified

low density cellulose fiber networks: A sustainable and fireproof alternative to

petroleum based foams Carbohydrate Polymers, 230, Article 115616

K¨oklükaya, O., Carosio, F., & Wågberg, L (2018) Tailoring flame-retardancy and

strength of papers via layer-by-layer treatment of cellulose fibers Cellulose, 25(4),

2691–2709

Leistner, M., Abu-Odeh, A A., Rohmer, S C., & Grunlan, J C (2015) Water-based chitosan/melamine polyphosphate multilayer nanocoating that extinguishes fire on

polyester-cotton fabric Carbohydrate Polymers, 130, 227–232

Li, H., Kruteva, M., Mystek, K., Dulle, M., Ji, W., Pettersson, T., & Wågberg, L (2020) Macro- and microstructural evolution during drying of regenerated cellulose beads

ACS Nano, 14(6), 6774–6784

Li, Y C., Mannen, S., Morgan, A B., Chang, S., Yang, Y H., Condon, B., & Grunlan, J C (2011) Intumescent all-polymer multilayer nanocoating capable of extinguishing

flame on fabric Advanced Materials, 23(34), 3926–3931

Li, Y C., Schulz, J., & Grunlan, J C (2009) Polyelectrolyte/nanosilicate thin-film

assemblies: Influence of pH on growth, mechanical behavior, and flammability ACS

Applied Materials & Interfaces, 1(10), 2338–2347

Maddalena, L., Carosio, F., Gomez, J., Saracco, G., & Fina, A (2018) Layer-by-layer assembly of efficient flame retardant coatings based on high aspect ratio graphene

oxide and chitosan capable of preventing ignition of PU foam Polymer Degradation

and Stability, 152, 1–9

Mateos, A J., Cain, A A., & Grunlan, J C (2014) Large-scale continuous immersion system for layer-by-layer deposition of flame retardant and conductive nanocoatings

on fabric Industrial & Engineering Chemistry Research, 53(15), 6409–6416

McCormick, C L (1981) Novel cellulose solutions.

Nolte, A J., Rubner, M F., & Cohen, R E (2005) Determining the Young’s Modulus of polyelectrolyte multilayer films via stress-induced mechanical buckling instabilities

Macromolecules, 38(13), 5367–5370

Notley, S M (2008) Effect of introduced charge in cellulose gels on surface interactions

and the adsorption of highly charged cationic polyelectrolytes Journal of the

Chemical Society Faraday Transactions, 10(13), 1819–1825

Oliveira, W D., & Glasser, W G (1996) Hydrogels from polysaccharides I Cellulose

beads for chromatographic support Journal of Applied Polymer Science, 60(1), 63–73

Osman, Z., & Arof, A K (2003) FTIR studies of chitosan acetate based polymer

electrolytes Electrochimica Acta, 48(8), 993–999

Picart, C., Lavalle, P., Hubert, P., Cuisinier, F., Decher, G., Schaaf, P., & Voegel, J.-C (2001) Buildup mechanism for poly (L-lysine)/hyaluronic acid films onto a solid

surface Langmuir, 17(23), 7414–7424

Picart, C., Mutterer, J., Richert, L., Luo, Y., Prestwich, G., Schaaf, P., & Lavalle, P (2002) Molecular basis for the explanation of the exponential growth of polyelectrolyte

multilayers Proceedings of the National Academy of Sciences, 99(20), 12531–12535

Rodahl, M., H¨o¨ok, F., Krozer, A., Brzezinski, P., & Kasemo, B (1995) Quartz crystal microbalance setup for frequency and Q-factor measurements in gaseous and liquid

environments The Review of Scientific Instruments, 66(7), 3924–3930

Schweizer, E (1857) Das kupferoxyd-ammoniak, ein aufl¨osungsmittel für die

pflanzenfaser Advanced Synthesis & Catalysis, 72(1), 109–111

Srikulkit, K., Iamsamai, C., & Dubas, S T (2006) Development of flame retardant polyphosphoric acid coating based on the polyelectrolyte multilayers technique

Journal of Metals, Materials and Minerals, 16(2), 41–45

Stafford, C M., Harrison, C., Beers, K L., Karim, A., Amis, E J., VanLandingham, M R., & Simonyi, E E (2004) A buckling-based metrology for measuring the elastic moduli

of polymeric thin films Nature Materials, 3, 545

Vuoriluoto, M., Orelma, H., Johansson, L.-S., Zhu, B., Poutanen, M., Walther, A., & Rojas, O J (2015) Effect of molecular architecture of PDMAEMA–POEGMA random and block copolymers on their adsorption on regenerated and anionic nanocelluloses

and evidence of interfacial water expulsion The Journal of Physical Chemistry B, 119

(49), 15275–15286

Wågberg, L., & Bj¨orklund, M (1993) Adsorption of cationic potato starch on cellulosic

fibres Nordic Pulp and Paper Research Journal (Sweden), 8, 399

Wågberg, L., Decher, G., Norgren, M., Lindstr¨om, T., Ankerfors, M., & Axn¨as, K (2008) The build-up of polyelectrolyte multilayers of microfibrillated cellulose and cationic

polyelectrolytes Langmuir, 24(3), 784–795

Wang, S., Lu, A., & Zhang, L (2016) Recent advances in regenerated cellulose materials

Progress in Polymer Science, 53, 169–206

Wang, Z., Hauser, P J., Laine, J., & Rojas, O J (2011) Multilayers of low charge density

polyelectrolytes on thin films of carboxymethylated and cationic cellulose Journal of

Adhesion Science and Technology, 25(6–7), 643–660

Wendler, F., Schulze, T., Ciechanska, D., Wesolowska, E., Wawro, D., Meister, F., &

Liebner, F (2012) Cellulose products from solutions: Film, fibres and aerogels The

european polysaccharide network of excellence (EPNOE) (pp 153–185) Springer

Woodings, C (2003) Fibers, regenerated cellulose Kirk-Othmer encyclopedia of chemical

technology, 11, 246–285

Xie, A F., & Granick, S (2002) Local electrostatics within a polyelectrolyte multilayer

with embedded weak polyelectrolyte Macromolecules, 35(5), 1805–1813