Estolide Molecular Weight Distribution Via Gel Permeation

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

J Am Oil Chem Soc (2019)

DOI 10.1002/aocs.12165

ORIGINAL ARTICLE

Estolide Molecular Weight Distribution via Gel Permeation


Chromatography
Grigor B. Bantchev1 · Steven C. Cermak1 · Amber L. Durham1 · Neil P. J. Price2

Received: 30 May 2018 / Revised: 28 September 2018 / Accepted: 1 October 2018


Published 2019. This article is a U.S. Government work and is in the public domain in the USA

Abstract Using the universal calibration and the Mark- approximately three times more sensitive than the currently
Houwink equation (MHE) (½ηj ¼ KMjα ), three batches of used methods for determination of EN values.
oleic estolide acids and their corresponding 2-ethylhexyl
esters were characterized using gel permeation chromatog- Keywords Estolides  Molecular weight distribution  Gel
raphy (GPC). The MHE parameters in tetrahydrofuran permeation chromatography  Intrinsic viscosity
(THF) at 40  C were determined (for acids: α =
0.442  0.003 and log10K=2.505  0.007, for esters: α J Am Oil Chem Soc (2019).
=0.531  0.006 and log10K =2.794  0.018). The fits of
the GPC chromatograms yielded also the oligomeric com-
position of the estolides, which can be used to calculate the Introduction
estolide number (EN) of an estolide mixture, and other
molecular-weight distribution parameters, such as number- Estolides are a class of bio-based oligomeric compounds,
average molecular weight (Mn), weight-average molecular which have shown excellent properties as lubricants
weight (Mw), and dispersity (Ð). Using the Deming line fit, (Cermak et al., 2017). Recently, estolides have been used
we concluded that the GPC should be expected to be as a base oil to formulate a series of bio-based engine oils
(5W-20 and 5W-30), which has been certified by the Amer-
ican Petroleum Institute (API). Biodegradability tests on
Supporting information Additional supporting information may be the formulated estolide motor oils have shown that the esto-
found online in the Supporting Information section at the end of the lide base oils maintain a high degree of biodegradability
article. even while formulated or blended with a host of different
additives. The estolide-based engine oils were tested in a
* Grigor B. Bantchev series of Nevada taxi cabs in an automotive engine for
grigor.bantchev@usda.gov thousands of miles and maintained their high level of bio-
1 degradability (Bredsguard et al., 2016).
Bio-oils Research Unit, National Center for Agricultural
Utilization Research, Agricultural Research Service, United Estolides are formed when the carboxylic acid function-
States Department of Agriculture, 1815 N. University Street, ality of one fatty acid (FA) reacts at the olefinic site of
Peoria, IL 61604, USA another FA to form a secondary ester linkage. The number
2
Renewable Product Technology Research Unit, National Center of these ester linkages represents the estolide number (EN).
for Agricultural Utilization Research, Agricultural Research The EN is one of the best means to describe the degree of
Service, United States Department of Agriculture, 1815
N. University Street, Peoria, IL 61604, USA oligomerization of an estolide. The EN can be determined
using gas chromatographic (transesterification-GC) analysis
Mention of trade names or commercial products in this publication is of the fatty acid methyl ester (FAME) composition after
solely for the purpose of providing specific information and does not
imply recommendation or endorsement by the U.S. Department of
methanolysis of the estolide, or by nuclear magnetic reso-
Agriculture. USDA is an equal opportunity provider and employer. nance (Cermak and Isbell, 2001). The EN is defined as the

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

the molecules are fractionated by their viscometric hydro-


dynamic size. The smaller molecules enter the pores of the
column resin, and this way they are retained, which leads
to their elution after the bigger molecules that cannot enter
the pores (Grubisic et al., 1967).
In the current article, we report the preparation of several
samples of oleic estolide acids (estolide acids) and their
corresponding oleic estolide 2-ethylhexyl esters (estolide
esters). The estolides differ by the synthesis reaction time
and the extent of distillation/purification, i.e., fractionation
by molecular weight (MW). The samples were analyzed by
GPC to determine the MW distribution and the average
MW. The limiting viscosity numbers (also known as intrin-
sic viscosity) for the oligomers in tetrahydrofuran (THF)
were also calculated, which can be used in future GPC
Scheme 1 Synthesis of estolide acids and estolide esters. Estolide evaluations of estolides. The EN was calculated from the
number (EN) of a molecule is equal to j−1 where j is the number of
starting FA. The estolide position was distributed from positions 5–13 number-average MW (Mn) and compared to the EN, deter-
(0 ≤ n ≤ 8) with the original Δ9 and Δ10 positions (n = 4 and 5) hav- mined from the transesterification-GC method.
ing the greatest abundances

average number of FA added to the base FA (Scheme 1, Experimental Procedures


EN = j−1, where j is the total number of FA moieties in
the molecule). For example, j = 2 and EN = 1 would repre- Materials
sent a dimer estolide ester (Scheme 2) or two FA linked
together with one ester linkage. EN values of two or more Hexanes, sodium sulfate, sodium chloride, sodium phos-
indicate estolides with multiple ester linkages and are hence phate monobasic monohydrate, THF, and toluene were
from Fisher Scientific (Pittsburgh, PA, USA). Butylated
called trimeric-plus estolides. The secondary ester linkages
hydroxytoluene (BHT), methyl palmitate (99%) (MePa),
of the estolide, resulting in a unique estolide structure, are
oleic acid (90%), perchloric acid (70%), and tristearin
reported to be more resistant to hydrolysis than those found
(99%) (TriSt) were from Sigma-Aldrich (St. Louis, MO,
in triacylglycerols, which yield materials with far superior
USA). Methanol, ethyl acetate, and potassium hydroxide
physical properties (Cermak and Isbell, 2003).
were purchased from EMD Millipore Co. (Billerica, MA,
During the synthesis of estolides, the EN can easily be
USA). Filter paper was obtained from Whatman (Clifton,
altered or manipulated by changing any of the following
NJ, USA). 2-ethylhexyl alcohol (2-EH alcohol) was pur-
synthetic parameters: temperature, time, catalyst type,
chased from Univar (Downers Grove, IL, USA). Sulfuric
and/or catalyst amounts. The ability to synthesize estolides
acid was purchased from LabChem Inc. (Zelienople, PA,
with a wide variety of EN has allowed researchers (Cermak
USA). Ethanol was purchased from AAPER Alcohol and
and Isbell, 2004) to discover predictable relationships
Chemical Company (Shelbyville, KY, USA). The FAME
between EN and certain estolide properties (pour and cloud
standard mixtures were obtained from Nu-Check Prep
points, viscosity, density, and/or wear/lubrication).
(Elysian, MN, USA). Solvents for chromatography and
Gel permeation chromatography (GPC), also known as
extraction were HPLC grade or an equivalent, and were
size-exclusion chromatography, is a well-known method
used without further purification. Universal calibration
for studying polymers and oligomers. Using this technique,
curve was prepared with EasiVial PS-L low-MW poly-
styrene standards from Agilent Technologies (Santa
Clara, CA, USA). The standards contained polystyrene
samples with peak MW (Mp) from 162 to 45,120 Da and
the log10(Mp[η]) between 0.42 and 4.0, where [η] was the
limiting viscosity number in dL g−1 (See the SI). The [η]
values of the polystyrene standards were supplied by the
manufacturer, except for the lowest MW (162 Da), which
Scheme 2 2EH ester of oleic dimer estolide (EN = 1, was extrapolated from the higher MW values, using
j = 2). 0 ≤ n ≤ 8 the Mark-Houwink equation (MHE).

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

Estolide Synthesis

Simplified Scheme of the Procedures and Nomenclature ni j

(a) 13-ni
Scheme 3 represents a simplified information flowsheet of
the synthesis and the distillation procedures used to obtain
the different estolide batches and fractions. It also provides
(b)
the nomenclature that will be used throughout the manu-
script. Full details of the synthesis and the distillation pro-
cedures are given in the next sections. During the course of
the estolide acid synthesis, it is possible that some cyclic (c)
estolides are formed (Scheme 4). While dimers and higher
cyclic estolides from this kind of synthesis have never been
reported, the cyclic monomers: delta and gamma lactones
(Scheme 4 b and c) were reported previously by Cermak (d)
and Isbell (2001). The lactones were detected in the GPC
analysis of samples: Eb, Ec, E1b, and E1c. Scheme 4 Cyclic estolides. (a) general formula. i = 1 to j, (b) gamma
lactone (j = 1, n1 = 1 in figure 4a), (c) delta lactone (j = 1, n1 = 2),
and (d) cyclic dimer estolide (j= 2, n1 = 5, n2 = 8)
Oleic Estolide Acids (Estolide Acids)

An acid-catalyzed condensation reaction was conducted (7.5–10.9 kPa), and stirred with a Teflon-coated stir bar.
without solvent in a 500 mL round-bottom flask that had Once the desired temperature of 60  0.1  C was reached,
been pretreated with acidic wash. Oleic acid (100.2 g, perchloric acid (17.74 mmol, 1.5 mL, 0.05 eq.) was added,
354.7 mmol) was heated to 60  C under house vacuum vacuum restored, and stirred. After the corresponding time

Oleic H+ catalyst, ∆ Estolide acids “Ax”


acid Quenching the catalyst “Ax”= “Aa” (24h), “Ab” (96h) or “Ac”(168h)

Estolide Removal of monomer Estolide


acids “Ax” acid by distillation acids “A –1 x”

Estolide Esterification with 2EH alcohol, Estolide esters


acids “Ax” Quenching the catalyst “Ex”

Removal of monomer
ester by distillation

Monomer Estolide esters


esters “E1x” “E–1x”

Removal of dimer
ester by distillation

Dimer esters Estolide esters


“E2x” “E-2x”

Scheme 3 Simplified synthetic path and nomenclature in the current manuscript. “x,” depending on the starting estolide acid mixture, stands for
“a,” “b,” or “c” (see top of the Scheme). Capital “A” stands for estolide acid, capital “E” stands for estolide ester. “1” stands for mostly the mono-
mer mixture, “−1” for mixture with the removed monomer, “2” for mostly the dimer mixture, “−2”—for trimeric-plus estolide esters. “Ax” sam-
ples were not tested by GPC

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

of reaction, i.e., 24, 96, or 168 h, the solution was allowed MW Fractionation—Molecular Distillations
to cool to room temperature with stirring. The reaction was
then quenched by the addition of KOH (1.144 g, Fractional distillations from the previous experiment were
20.40 mmol, 1.15 eq. based on HClO4) in 90% ethanol/ performed on the resulting set of estolide esters (E−1a,
water (20 mL) solution and allowed to stir for at least E−1b, and E−1c) using a Myers Lab 3 short-path molecular
30 min. The mixture was allowed to settle and was filtered distillation unit (Myers Vacuum, Kittanning, PA, USA).
through a Buchner funnel using Whatman #54 filter paper. This step of distillation was performed to remove the dimer
The pH of the solution was adjusted to 5–6 using a pH 5 estolide 2-EH esters (Dimer Esters, EN = 1) from the esto-
buffer (1 M NaH2PO4). The organic layer was then washed lide ester mixture leaving behind the larger oligomers
with saturated sodium chloride solution, dried over sodium known as trimeric-plus estolides. Cermak et al. (2012) pro-
sulfate, and filtered with Whatman #54 filter paper. All vided in-depth information on the capabilities, distillation
reaction products were Kugelrohr-distilled at 180–200  C diagram, and exact parameters for fractional distillation
at 0.013–0.067 kPa to remove any unreacted FA and by- using the Myers Lab 3 unit. For this distillation, the con-
products, such as lactones (Cermak and Isbell, 2001). The denser temperature was set at 30  C and cold tap water was
resulting estolide acid products (named A−1a, A−1b, and used to cool the diffusion pump and rotor bearing. Vacuum
A−1c) were then filtered with Whatman #54 filter paper. was maintained at 1.60–3.73 Pa at both the chamber and
foreline pressure sensors. Rotor temperature was main-
tained between 224 and 228  C and the feed was adjusted
Oleic Estolide 2-Ethylhexyl Esters (Estolide Esters) such that the estolide materials entered the rotor chamber at
about 40 drops min−1. Feedstock was fed onto a heated
Reactions were conducted without solvent in a 500 mL rotor spinning at 28.75 Hz under a high vacuum
round-bottom flask that had been pretreated with acidic (1.60–3.73 Pa).
wash. Oleic acid (200.1 g, 708.3 mmol) was heated to As the distillation proceeded, the feedstock of estolide
60  C under house vacuum (7.5–10.9 kPa), and stirred esters (E−1a, E−1b, and E−1c) was fractionated to yield a
with a Teflon-coated stir bar. Once the desired tempera- mostly dimeric portion of the estolide esters (EN = 1) in
ture of 60  0.1  C was reached, perchloric acid the distillate (dimer esters: E2a, E2b, and E2c) and a
(35.41 mmol, 3.1 mL, 0.05 eq.) was added, vacuum trimeric-plus oligomer portion in the residue (EN >1) (tri-
restored, and stirred. After the corresponding time of meric-plus estolide esters: E−2a, E−2b, and E−2c). The
reaction, i.e., 24, 96, or 168 h, 2-EH alcohol (110.7 g, residual trimeric-plus estolide portion was a mixture of tri-
849.9 mmol, 133.0 mL) was added to the vessel, vacuum meric and larger estolide esters, including a smaller portion
was restored, and the mixture was stirred for an addi- of undistilled dimeric estolides. The short residence time at
tional 6 h. The completed reactions were quenched by high temperature of the Myers distillation unit helps reduce
the addition of KOH (2.285 g, 40.7 mmol, 1.15 eq. degradation and color bodies of the residue unlike other
based on HClO4) in 90% ethanol/water (20 mL) solution distillations where the sample remains at the distillation
and allowed to stir for at least 30 min. After settling, the temperature for several hours. The short residence time and
mixture was filtered through a Buchner funnel using the high vacuum also make it possible for higher distilla-
Whatman #54 filter paper. The pH of the solution was tion temperatures, which are necessary for removal of the
adjusted to 5–6 using a pH 5 buffer (1 M NaH2PO4). heavy dimeric estolides.
The organic layer was then washed with saturated
sodium chloride solution, dried over sodium sulfate, and Transesterification-GC–MS Method for the
filtered with Whatman #54 filter paper. All reaction prod- Determination of the EN
uct mixtures were Kugelrohr distilled at 90–110  C at
0.013–0.067 kPa to remove any excess 2-EH alcohol. At The EN, which identifies the extent of oligomerization of
this point, the estolide ester samples, Ea, Eb, and Ec, the product, is defined as the average number of FA
were collected. The remaining products were further dis- added to the base FA. Upon methanolysis of an estolide
tilled by Kugelrohr distillation at 180–200  C at sample with an EN = 2 ( j = 3, Scheme 1), the product
0.013–0.067 kPa to remove any unreacted FA and by- will have components: two hydroxy FAME and one
products, such as lactones (Cermak and Isbell, 2001) and nonhydroxy FAME.
oleic 2-EH esters (Monomer Esters, EN = 0). The final The EN is calculated from the transesterification-GC
products were filtered using Whatman #54 filter paper. method where the ratio of the areas is measured, i.e., the
This final distillation resulted in the collection of estolide hydroxy FAME peaks to the nonhydroxy FAME peaks as
esters (E−1a, E−1b, and E−1c) in the residue and Mono- described previously (Isbell and Kleiman, 1994). In the
mer Esters (E1a, E1b, and E1c) in the distillate. example above, the ratio will be 2:1.

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

GC–MS Characterization of Low Estolides group being a part of the estolide chain, as opposed to an
end group, is the same for all of them. This assumption
The estolide esters with low MW (monomers and dimers) leads us to the conclusion that during the synthesis of
were run without pretreatment (transesterification) on a estolide acids there is a geometric distribution of the olig-
SPB-1 column. The temperature profile was a starting tem- omeric species:
perature of 150  C, a ramp of 4  C min−1 for 40 min  
Mj + 1
(to 310  C), followed by a hold for 40 min at 310  C. The   ¼p ð1Þ
helium flow rate was 1 mL min−1, the injection volume Mj
was 2 μL, and the split ratio was 10:1. The sample concen- for every j ≥ 1, where 0 ≤ p < 1. j is the number of
tration was ~10 mg mL−1. monomeric units in the oligomer, the square brackets in this
case refer to the molar concentration of the species. This is
Matrix-Assisted Laser Desorption/Ionization Time-of- known as a Flory MW distribution (Kissin, 1995). The
Flight MS (MALDI MS) Flory’s distribution is expected to be preserved during the
esterification of the estolide acids to estolide esters, but it
MALDI MS spectra were obtained on a Bruker-Daltonic will be distorted during the distillations due to the higher
Microflex (Billerica, MA, USA) instrument operating in the volatility of the monomers and lower oligomers. As we did
reflectron mode, with lens and reflector voltages of 9.20 and not detect (by GPC) any higher than tetramer oligomers to
20.00 kV, respectively. Laser excitation (337.1 nm) was distill, we assumed that the studied estolides obey the
typically at 75% of 150 μJ maximum output, and 1200 shots Flory’s distribution for pentamers and higher oligomers.
were acquired. The matrix use was 2, 5-dihydrobenzoic acid The concentrations of the monomers to tetramers were
(2, 5-DHB). fitted individually, even in the cases (Ea, Eb, and Ec) where
all the oligomer concentrations were expected to obey the
GPC Runs Flory’s distribution (Eq. 1).
Jacobson and Stockmayer (1950) published a theory
The unfiltered samples were run on a GPC consisting of a (later refined by Flory and Semlyen, 1966) for the concen-
precolumn PLgel 3 μm Guard (50 × 7.5 mm, Part tration of cyclic (ring) polyesters. The Flory distribution for
1110–1320), two PLgel 3 μm mixed E columns linear polyesters (Eq. 1) can be written as:
(300 × 7.5 mm) from Agilent Technologies (Santa Clara,  
CA, USA) connected in a series, a Wyatt Optilab REX dif- Mj ¼ Bp j ð2Þ
ferential refractive index (dRI) detector, a Wyatt Treos
where B is a constant. The Jacobson-Stockmayer distri-
miniDAWN light-scattering detector from Wyatt Technol-
bution can be written as:
ogy (Santa Barbara, CA, USA), a Waters 717 autosampler,
  CVp j .
and a Waters 1515 isocratic pump from Waters Corporation Rj ¼ ð3Þ
(Milford, MA, USA). Forty to fifty milligrams of a sample 2:5 j

were dissolved in 1 mL of THF with 0.1% BHT (stabilizer) where Rj is a cyclic estolide with j FA, C is a constant,
and 0.5% toluene (internal standard for the flow rate). Half and V is the volume of the system. From these distribu-
of the estolide samples were prepared in duplicate on a sep- tions, it can be derived that the ratio of cyclic estolides with
arate day and rerun to determine the reproducibility of the j FA to the linear estolides with the same number of FA is:
data. The EasiVial standards were prepared by dissolving . .
the content of the vial in 2 mL THF with 0. 1% BHT and ½Rj  ¼ CV=B 2:5 ð4Þ
0.5% toluene. Hundred microliters were injected and the ½M j  j

solvent (THF with 0.1% BHT) was run at 1 mL min−1. and


The columns were kept at 40  C during the runs. In addi-
  
tion to the estolide and calibration samples, MePa and TriSt     Mj + 1 j 2:5
samples were run as monodisperse lipid samples. Rj + 1 ¼ Rj   ð5Þ
Mj j+1

Theoretical The Jacobson and Stockmayer theory does not take


into account the ring strain, and work on cyclic oligo-
Oligomer Distribution mers has shown that the cyclic oligomers with rings of
8 to 12 atoms to be strained (Illuminati and Mandolini,
If we assume that the carboxylic head of a FA is separate 1981). Correspondingly, the Eqs 3–5 are valid only for
from the rest of the estolide molecule and hence not influ- j ≥ 2 and cyclic monomers (lactones, j = 1) do not con-
enced by it, then the probability, p, of the carboxylic form them. There is an example in the literature that

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

   
rings with 9 to 13 atoms form in a lesser amount than zpj ¼ log10 Mj ½ηj ¼ log10 KMjα + 1
the corresponding dimer cycles (Ruddick et al., 1999), 
due to the straining. ¼ log10 ðK Þ + ðα + 1Þlog10 Mj ð12Þ

This formula connects peak elution volumes with the


Universal Calibration oligomer molar mass and two parameters, specific for the
oligomeric series and the experimental conditions (solvent
The principle of the universal calibration states that all and temperature). It should not depend on the GPC column,
polymers with the same value of the product M[η] have the as long as the oligomers are within its range.
same elution volumes (Grubisic et al., 1967):
log10 ðM ½ηÞanalyte ¼ log10 ðM ½ηÞcalibration standard ð6Þ Computational

when Preparation of Chromatograms


Vanalyte ¼ Vcalibration standard ð7Þ
The instrument was equipped with a light-scattering detec-
In the above statement M is the MW, [η] is the limiting tor, but due to the relatively low MW of most of the esto-
viscosity number (also known as the intrinsic viscosity), lides, it did not produce usable signals. The dRI was
and V is the peak volume at which the material elutes from monitored as a function of the elution volume. All the com-
the GPC column. Using the universal calibration is possible pounds at these instrument settings eluted after 12 mL, so
if [η] of the molecules is known. Sometimes when [η] data the signal between 1 and 10 mL was used to calculate a
are not available, a simplification is made: baseline as an average value of the dRI signal. The baseline
log10 ðM Þanalyte ¼ log10 ðM Þcalibration standard ð8Þ was subtracted, and the chromatogram was smoothed with
a Savitzky–Golay filter of second order (Savitzky and
when Golay, 1964). The optimal window size for the smoothing
(in most cases 11 datapoints, spanning ~0.096 mL, in one
Vanalyte ¼ Vcalibration standard ð9Þ
case 7 datapoints) was selected using the method described
This approach, called “GPC calibration,” works when by Vivó–Truyols and Schoenmakers (2006).
the oligomers used for the calibration are similar to the ana- The calibration curve was determined to be
lyzed ones. It is shown that using polystyrene standards for
z  log10 ðM ½ηÞ ¼ −0:0039093328V 3 + 0:18410884V 2
the simpler calibration leads to wildly higher than expected
MW of lipid samples (Bantchev et al., 2016; Noor et al., −3:296576V + 23:009613
2016). This is due to the large dissimilarities between the ð13Þ
two kinds of molecules. Universal calibration, which
where V is the elution volume in mL.
requires knowledge of [η], is needed.
The chromatogram was recalculated to equally space
To determine the [η] values for a series of homologous
z ≡ log10(M[η]) from 3.3 to 0.74 axis with a step of 0.002
oligomers, the Mark-Houwink Equation (MHE) is used:
(11.8 to 18.2 mL, step 0.0034 to 0.005 mL). The dRI was
½ηj ¼ KMjα ð10Þ dz
recalculated inversely proportionally to the dv derivative
where j is the number of monomer units in the oligomer, dz
− ¼ 3 × 0:0039093328v2 − 2 × 0:18410884v + 3:296576
K and α are constants, specific for each oligomeric series, dv
and Mj is the MW of the oligomer with j units. ð14Þ
Using well-characterized polystyrene standards, a uni-
versal calibration curve can be made: The curve fitting was performed in recalculated dRI ver-
sus z coordinates.
z  log10 ðM ½ηÞ ¼ f ðVelutionÞ ð11Þ

It is quite typical for z to be a linear function of V. In our Curve Fitting of Individual Peaks
case, we observed a slight nonlinear trend at higher MW,
so a cubic polynomial was used instead. zpj (the peak elu- In a preliminary fitting, four fitting curves were tested:
tion volumes for estolide oligomers with j units) were Gaussian (three parameters per peak), log-normal (three
determined from GPC chromatograms (see the “curve- parameters), Poisson (four parameters), and generalized
fitting of whole chromatograms” section). exponentially modified Gaussian (EMG) (four parameters
Using the MHE and the universal calibration equation, per peak) (Di Marco and Bombi, 2001). The Gaussian and
the following formula can be derived: the EMG curves were better fits. The Gaussian had the

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

advantage of being simpler and allowing easier calculations Neglecting the higher oligomers (j > 20) was estimated to
of the curve characteristics (position of the peak maximum lead to an error of less than 0.02% of the total area.
and area under the curve) from the fitting parameters, while The three samples (A−1a, A−1b, and A−1c) of estolide
EMG gave a better fit due to the present asymmetry of the acids were run in duplicate. These runs yielded six chro-
peaks. Finally, the EMG was selected as the best fit for the matographic curves that were fitted simultaneously with
data presented here. 4 + 5*6 = 34 fitting parameters. The estolide esters ana-
The equation for the generalized EMG is (Di Marco and lyzed were the undistilled esters (Ea, Eb, and Ec)
Bombi, 2001): (in duplicate), the esters after the removal of the monomers
rffiffiffi  2  
s σ z −μ
(E−1a, E−1b, and E−1c), and the esters after the partial
hσ π σ z −μ
y¼ exp − erfc pffiffiffi − ð15Þ removal of dimers (E−2a, E−2b, and E−2c) (in duplicate).
jτj 2 2τ2 τ 2 τ σ There were a total of 15 chromatograms fitted simulta-
where μ is the position of the unmodified Gaussian peak, neously with 4 + 5*15 = 79 fitting parameters.
σ is the Gaussian width of the peak, τ is the relaxation time, The jackknife method (Caceci, 1989) was used to deter-
related to the tailing of the peak, h is the Gaussian height of mine the average and the standard deviation of the common
the peak, π is the constant 3.14…, erfc() is the complemen- fitting parameters: log10(K), (α + 1), σ and τ. log10(K) and
tary error function, z ≡ log10(M[η]) is the abscissa value, α were used with the MHE to calculate [η]j of the estolide
and s is =1 if τ >0 and = − 1 if τ <0. oligomers. Using the jackknife method, the averages and
The generalized EMG differs from EMG in that negative the standard deviations are determined by fitting data sub-
values for τ are acceptable, which allows the generalized sets with an omitted datapoint, and averaging the resulting
EMG to represent fronted peaks. All the peaks were tailed fitting parameters. In the current implementation of the
in the representation of the signal as a function of the eluted jackknife method, a GPC chromatogram was treated as a
volume, which corresponded to fronted peaks when the sig- datapoint.
nal was presented as a function of the z (≡log10(M[η]) coor- The averages and the standard deviations of the oligo-
dinate, due to the negative slope between the two. meric composition and related parameters were calculated
A computational complication arose from the fact that from the values for the duplicate runs.
the calibration was made based on the position of the maxi-
mums of the peaks, zp, while the fitting procedure did not
yield them directly. The formula to calculate them is not Calculation of the MW Averages
explicit (see Appendix A).
The following formulas were used to calculate the MW
averages:
Curve Fitting of Whole Chromatograms P 
w =M − 1
Mn ¼ Pi i ð16Þ
wi
Chromatograms were fitted using the EMG curves for each
P
individual peak. The positions of the estolide peaks were wi Mi
assumed to obey the universal calibration and the MHE Mw ¼ P ð17Þ
wi
(Eq. 12). log10(K) and (α + 1) were fitting parameters.
Assuming that most of the width of the peaks was due to where Mn and Mw are the number-average and weight-
band broadening, we used the same values for σ and τ for average MW. The Eqs 16 and 17 can be used in two ways.
all the peaks in the fit. There are studies that have shown In the first, Mi refers to the MW of the ith estolide oligo-
weak or no dependence of these parameters on the elution mer, and wi is the mass amount of this oligomer (propor-
position of the peak (Busnel et al., 2001) when the MW is tional to the area of the peak). It can be used when the
not close to the cutoff limits of the column. composition is known. In the current manuscript, the com-
Several chromatographic curves were fitted simulta- position is determined from fitting of the chromatograms.
neously, using the same log10(K), (α + 1), σ, and τ as fitting The results will be denoted with a subscript “GPC1” if
parameters. Five more fitting parameters were used per needed.
chromatogram, and they were related to the area under each In the second application, the chromatographic data are
peak (which is proportional to the concentrations of the used. Then, wi is the dRI signal at slice i, and Mi is deter-
oligomer). As mentioned in the “oligomeric distribution” mined from the universal calibration and the MHE:
section, the areas, Aj, of the first four peaks (1 ≤ j ≤ 4) were pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Mi ¼ α + 1 ð10zi =K Þ ð18Þ
fitted as independent parameters, while for the rest
(5 ≤ j ≤ 20) a fitting parameter p and the equation Aj = pAj − 1 In the above equation, zi is calculated from the calibra-
were used according to the Flory’s distribution (Eq. 1). tion curve (Eq. 13). In this approach, if log10(K) and

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

(α + 1) are known, the MW averages can be calculated z=log(M[η])


3.0 2.5 2.0 1.5 1.0 0.5
without curve fitting the chromatogram. The summations
were carried out for z values of 0.9 to 3.3 for the estolide 4
10
acids and 0.956 to 3.3 for the estolide esters. The signal for

peak MW, Da
dRI, (a.u.)
z > 3.3 was essentially equivalent to the baseline. The sig-
3
nal for the low z values contained, in some cases, a lactone A-1b 10
Eb
peak that we wanted to treat separately (see results and dis- E-2b

cussion). In the current manuscript, the results obtained 2


10
using this approach will be indicated with a subscript
12 13 14 15 16 17 18
“GPC2,” if needed. Retention volume, (ml)
Other related parameters are dispersity (Ð) and EN:
Mw Fig. 1 Example of GPC chromatograms of estolides: differential
Đ¼ ð19Þ refractive index as a function of elution volume. Solid line: Estolide
Mn
acids formed after 96 h of synthesis (A−1b). Dotted line: the undis-
Ð is = 1 for monodisperse samples and >1 for polydis- tilled estolide esters (Eb). Dashed line: the estolide esters, after
removal of monomers and partial removal of dimers by distillation
perse samples. (E−2b). The markers show the Mp of the standards used to create the
Mn −M0 universal calibration curve (right axis) vs. the peak retention volume
EN ¼ −1 ð20Þ (bottom axis) or z (top axis)
ΔM
The EN is a measure of the average number of oligo- Estolide Acids
meric units. M0 is the MW of the end groups (= 0 for esto-
lide acids, = 112 for estolide esters). ΔM is the MW of a The chromatograms from the GPC analysis are presented in
monomeric unit (=282 for estolides based on oleic acid). Fig. 2. It shows relatively small differences between the
samples. Nevertheless, there was a reproducible result that
the longer reaction times led to a decrease in the relative
Software
proportion of the lower oligomers and increased the propor-
tions of the higher oligomers ( j > 5).
The calculations were performed with Microsoft Excel
The chromatograms were well fitted (see Fig. 3 for an
2013. Some of the calculations used Visual Basic for
example of the fit of one chromatogram and Table 1 for the
Applications. Curve fitting utilized the Solver Add-In to
common fitting parameters and the R2 value).
find the minimum of the sum of squared differences
The composition of the estolide acids, obtained from the
between the chromatograms and the fitting curves.
fitting of the GPC data, is given in Table 2. We can see that
lower oligomers ( j < 4) decrease with the length of the
reaction, while the higher oligomers ( j ≥ 5) increase.
Results and Discussion Table 2 also includes the MW averages and derived
parameters.
Representative Results
a
b
Fig. 1 shows three GPC chromatograms plotted as a func- c
tion of the elution volume, V. We can see that the peaks of
Recalculated dRI, a.u.

4–5 lower oligomers can be easily identified; the higher


a
oligomer peaks overlap. While the peak positions of the b
estolide esters from the Eb sample are close to the corre- c

sponding ones from the other sample (E−2b), the ones from
c
the estolide acids (A−1b) are at different elution volumes. b
a
This is because the same oligomers have the same elution
volume, irrespectively of the run, or the exact composition.
It can be seen that for Eb, undistilled estolide esters, the 1.0 1.5 2.0 2.5 3.0
log(M[η])
area of the monomer peak (highest elution volume) is
higher, and it decreases with the increase of the oligomer
number. Fig. 2 GPC chromatograms of estolide acids (A−1a, A−1b, and A−1c).
Two separate chromatograms for each sample are run and shown.
On the top of the graph, the z-axis is plotted. It can be
Solid line: Estolide acids synthesized for 1 day (A−1a). Dotted line—
seen that the distortions from the linearity are very minor— estolide acids synthesized for 4 days (A−1b). Dashed line—estolide
it is hard to notice the unequal spacing of the z tick marks. acids synthesized for 7 days (A−1c)

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

E–1b (Monomer removed)


Data (A–1b) Eb (un-distilled)
Overall fit

Recalculated dRI, a.u.


Recalculated dRI, a.u.

Fit, individual peaks

E–2b (Dimer reduced)


j=7 E–1b (Monomer removed)
j=1 j=2 j=3 j=4 j=5 j=9 j=11 Eb (Undistilled)
L j=1 j=2 j=3 j=4 j=5

1.0 1.5 2.0 2.5 3.0 1.0 1.5 2.0 2.5 3.0
log(M[η]) log(M[η])

Fig. 3 Example of GPC chromatogram of estolide acids formed after Fig. 4 GPC chromatograms of the estolide esters: Eb, E−1b, and E−2b
4 days of synthesis (A−1b) (solid line), fit (dashed line), and the individ- (solid lines), and their fits (dashed lines). “L” denotes the lactone
ual fitting curves for the first six odd-numbered oligomers (dotted lines) peak. Its range is excluded from the fit

Estolide Esters
initial estolide esters were fractionated by distillation to
Evaluation of the Polydisperse Samples separate them into monomer-rich fractions (E1a, E1b, and
E1c) and monomer-free estolide esters (E−1a, E−1b and
Estolide esters (Ea, Eb, and Ec) were synthesized from E−1c). The monomer-free estolide esters were distilled fur-
estolide acids, by esterification with 2-EH alcohol. The ther into dimer-rich fractions (E2a, E2b, and E2c) and
Table 1 Fitting parameters for the chromatogram

Estolide acids Estolide esters MePa TriSt

n 6 15 4 3
(α + 1) 1.442  0.003 1.531  0.006 n.a. n.a.
log10K 2.505  0.007 2.794  0.018 n.a. n.a.
σ 0.0404  0.0005 0.0412  0.0001 0.0416  0.0013 0.0401  0.0001
−τ 0.0476  0.0006 0.0440  0.0003 0.0359  0.0044 0.0317  0.0010
R2 0.9962 0.9925 0.9987 0.9996

The parameters are determined with an abscissa z = log10(M[η]), where M is in Daltons and [η] in dL g−1 (units of the abscissa log10( mole−1 )).
n.a.; not applicable to compounds that are not a part of an oligomeric series. R2 is the squared coefficient of determination of the fits

Table 2 Composition of estolide acids and some descriptors, determined from fits of the GPC chromatograms

A−1a A−1b A−1c


1 day 4 days 7 days

% monomers 1.04  0.02 0.67  0.05 0.40  0.01


% dimers 29.4  0.1 28.0  0.2 27.3  0.1
% trimers 22.8  0.1 22.2  0.2 21.8  0.1
% tetramers 15.36  0.08 15.38  0.08 15.28  0.06
% pentamers 10.33  0.02 10.57  0.02 10.67  0.01
% higher oligomers 21.1  0.3 23.1  0.5 24.5  0.3
p 0.672  0.003 0.687  0.004 0.698  0.003
Mn (GPC1) 880  2 907  5 925  3
Mw (GPC1) 1157  5 1202  11 1232  7
Ð: (GPC1) 1.315  0.003 1.325  0.004 1.333  0.003
EN (GPC1) 2.12  0.01 2.22  0.02 2.28  0.01
Mn (GPC2) 851  1 876  6 894  1
Mw (GPC2) 1161  7 1172  51 1231  6
Ð: (GPC2) 1.364  0.006 1.34  0.05 1.377  0.005
EN (GPC2) 2.017  0.004 2.107  0.022 2.169  0.004

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

trimeric-plus estolide esters with a reduced amount of dangling chains, which leads to their contraction. If this is
dimers (E−2a, E−2b, and E−2c). GPC chromatograms of the the case, THF should be even poorer quality solvent for the
estolide esters, obtained from 4-day-synthesized estolide estolides, because the low slope (α value) is observed at
acids (b-series), are shown in Fig. 4. The chromatograms of low MW.
the undistilled estolide esters (Ea, Eb, and Ec), formed from
different estolide acids, are shown in Fig. 5. Presence of Cyclic Estolides
The fitted compositions of the estolide esters are pre-
sented in Table 3. It can be seen that there is no statistically A potential complication is the possibility that cyclic esto-
significant difference between “b” esters and the corre- lides can form during the estolide acid synthesis
sponding “c” esters (the lactone was excluded from the (Scheme 4). The formation of cyclic oligomers is usually
calculations). favored when the starting concentration of the monomer
The log10 K, (α + 1), σ, and τ values, determined from (oleic acid) is low. Lactones were reported earlier, and we
the fit, are included in Table 1. The table also includes the have detected them in Eb and Ec samples by GPC. To
σ and τ values for the two monodisperse lipids—MePa and determine the possible presence of cyclic dimers, some
TriSt. The similarities between σ and τ values for estolide samples were evaluated with MALDI and GC–MS (without
acids, estolide esters, MePa, and TriSt suggest that the peak transesterification).
widths are due to a large extent to the band broadening The 2-EH esters could be identified in the GC–MS sam-
from the column, and to a lesser extent to the dispersity ples by the presence of a prominent fragment with m/z
within the oligomers (from oligomers with the same num- 112 (2-ethyl 1-hexene ion) or 113. There was a group of
ber of units j, but different values of n in Scheme 1, for weak peaks that lacked these fragments. The presence of
example). fragment ions >400 Da suggested that they were dimers.
The values of (α + 1) for estolide acid and esters are low They were observed in E2b and E2c samples at slightly
compared to the slope of poly(hydroformylated methyl oleate) above 4 area %.
(PHFMO), determined in THF at 30  C (Milic et al., 2012). The ions with m/z 587.85–587.88 are the expected dimer
PHFMO has a structure that is close to (but different from) estolide acids or dimer cyclic estolide (C36H68O4Na+
the structure of the estolides. Similar to estolides, PHFMO = 587.50) and are observed in the MALDI-MS data, where
has a polyester backbone and a dangling hydrocarbon they were a minor peak. Unfortunately, MALDI-MS
chain. The dangling hydrocarbon chains are n-octyl or methods cannot definitely distinguish between linear esto-
n-nonyl for PHFMO, while in the estolides they can have lide acids and cyclic estolides, because they are isomers
different lengths (the average side-chain lengths for the and have identical molecular mass.
estolides and PHFMO are expected to be close). PHFMO Taking Eq. 5 and the result from the GC–MS that the
also has an extra CH2 group in the backbone repeat
ratio ½R2  ½M2  is approximately 4 area %, we can estimate
unit, and the end groups are different (methyl vs. 2-EH).
The slope for PHFMO in the Milic et al.’s study was the concentration of all the cyclic estolides in the mix. Con-
close to α=0.7 for the lower MW polymers, and decreased centrations of estolide esters Eb and Ec were calculated to
at high MW. The authors explained the decrease of the be 1.6% and 1.9%, respectively. This evaluation was based
slope with the THF being a poor quality solvent for the on the assumption that the peaks in the mixtures with ~4
area % were cyclic dimer estolide, and not something else,
like dimer estolide acids that remained during the esterifica-
Ea (1 day) tion with 2-EH alcohol.
Eb (4 days) Considering that, we should expect that the cyclic esto-
Recalculated dRI, a.u.

Ec lides were present in a low amount and did not influence


significantly the α values of the mixtures.

L
Ec Ec (7 days) Evaluation of Monomers and Dimers
Eb Eb (4 days)
Ea j=1 Ea (1 day)
j=2 j=3 j=4 j=5 When we fitted the GPC chromatograms of the monomer
1.0 1.5 2.0 2.5 3.0
esters (E1a, E1b, and E1c) and dimer esters (E2a, E2b, and
log(M[η]) E2c) distilled away from the estolide mixture, we observed
that they contained more than 88% of the main component
(oleate 2-EH ester (EN = 0) or dimer estolide 2-EH ester
Fig. 5 GPC chromatograms of estolide esters Ea, Eb, and Ec (solid
lines), and the fit to Eb (dashed line). “L” denotes the lactone peak. Its (EN = 1)). This led to overloading the column, and the
range is excluded from the fit main peak was shifted relative to the corresponding one in

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

J Am Oil Chem Soc (2019)


Table 3 Composition and related parameters of estolide esters, determined from fits of the GPC chromatograms

Ea Eba Ecb E−1a E−1b E−1c E−2a E−2b E−2c


1 day 4 days 7 days 1 day 4 days 7 days 1 day 4 days 7 days

% monomers 33.0  0.0 25.1  0.0 25.2  0.0 0.6 0.7 0.5 0.4  0.0 0.4  0.0 0.3  0.0
% dimers 26.2  0.1 23.4  0.1 23.5  0.1 39.2 31.4 31.6 24.1  0.1 21.4  0.8 21.0  0.0
% trimers 16.8  0.1 17.2  0.1 17.1  0.1 25.1 23.1 23.1 30.2  0.0 26.1  0.3 26.2  0.1
% tetramers 9.87  0.02 11.65  0.01 11.60  0.03 14.5 15.3 15.3 18.27  0.02 17.91  0.34 18.06  0.02
% pentamers 5.82  0.04 7.69  0.03 7.67  0.04 8.5 10.1 10.1 10.90  0.02 11.75  0.17 11.86  0.00
% higher oligomers 8.4  0.1 14.9  0.2 14.9  0.2 12.1 19.3 19.4 16.1  0.1 22.4  0.1 22.6  0.1
P 0.589  0.002 0.660  0.002 0.661  0.001 0.588 0.658 0.659 0.597  0.002 0.656  0.003 0.656  0.001
Mn (GPC1) 642  1 727  1 726  1 920 998 1002 997  1 1056  5 1065  0
Mw (GPC1) 863  3 1034  4 1032  3 1084 1232 1235 1201  5 1253  2 1314  3
Ð: (GPC1) 1.344  0.003 1.421  0.003 1.422  0.003 1.178 1.235 1.233 1.205  0.004 1.187  0.004 1.234  0.003
EN (GPC1) 0.880  0.002 1.182  0.004 1.177  0.003 1.865 2.141 2.155 2.235  0.005 2.443  0.022 2.471  0.003
Mn (GPC2) 626  2 712  2 710  2 891 971 976 999  1 1058  5 1066  0
Mw (GPC2) 854  2 977  4 1016  2 1070 1168 1222 1202  5 1254  2 1315  3
Ð: (GPC2) 1.364  0.000 1.372  0.003 1.432  0.002 1.20 1.20 1.25 1.203  0.004 1.185  0.004 1.233  0.003
EN (GPC2) 0.824  0.006 1.129  0.006 1.119  0.008 1.761 2.046 2.065 2.145  0.004 2.353  0.018 2.384  0.001
a
There was a lactone peak with area of 1.6% of the oligomer content.
b
There was a lactone peak with area of 3.0% of the oligomer content (the lactone amounts are not included in the total).
J Am Oil Chem Soc

the other estolide mixtures toward lower z values (higher Table 5 Limiting viscosity numbers, [η], in THF at 40  C for some
elution volumes). This shift led to a poor fit, when a joint estolide oligomers. Values  standard deviation
fit was attempted (all the chromatograms with same param- j [η]j, 1000× dL g−1
eters for (α + 1), log10K, σ, and τ).
Acids 2-EH esters
To account for the effect of overloading the column,
chromatograms of the monomer esters and dimer esters 1 40.7  0.10 41.2  0.17
were fitted with the parameters (α + 1), log10K, σ, and τ, 2 55.3  0.22 54.9  0.17
obtained for the estolide esters (Table 2), but using an addi- 3 66.2  0.32 66.0  0.29
tional fitting parameter per chromatogram—the shifting 5 83.0  0.50 84.4  0.58
factor for the major peak. The results obtained from these 7 96.3  0.66 99.8  0.87
fits are listed in Table 4. While the use of a shifting factor 10 112.7  0.87 120  1.3
can be implemented in GPC1 calculations, it cannot be 15 135  1.2 147  2.0
implemented in the GPC2 calculations unless there is prior 20 153  1.5 171  2.6
research correlating with the loading of the column with
the shift factor. Because we applied the GPC2 calculation
determine their [η] as 0.035 and 0.031 dL g−1, respectively.
without correction for overloading, the GPC2 results for
The estolide acid monomer consists of oleic acids and their
the MW and EN are shifted toward lower values.
isomers (the acid catalyst changes the configuration and the
position of the double bond). Its [η] of 0.0407 dL g−1 is
Calculated Limiting Viscosity Values close to the value for the oleic acid (0.035 dL g−1). The
discrepancy between the values for TriSt (0.061 dL g−1)
The limiting viscosity numbers, [η], for the estolide oligo- and triolein (0.031 dL g−1) is much bigger, which indicates
mers can be calculated from the parameters in Table 1 and a very strong effect of the double bonds on the limiting vis-
the MHE (Eq. 10). As the errors in the determination of the cosity number. The discrepancy between the present and
constants (α + 1) and log10K are probably correlated, the previous research (Abidi and Warner, 2001) could be
Table 5 lists the values and the associated standard devia- due to the fact that the temperature of the GPC experiment
tion, derived using the jackknife method. in the referenced study is not specified.
In this study, we also used MePa and TriSt. We calcu- It should be noted that there are studies that cast in doubt
lated their [η] to be (26.4  0.1) × 0.001 and the principle of universal calibration for low MW polymers.
(61.0  0.2) × 0.001 dL g−1, respectively. The lactone, They have shown the universal calibration curves (M[η]
encountered in some of the estolide samples, had [η] = vs. elution volume) for three kinds of molecules (polysty-
(26.1  0.6) × 0.001 dL g−1. rene, polyisobutene, and n-alkanes) deviate in the low
We found some GPC data about oleic acid and triolein, (<1000 Da) MW range (Chance et al., 1995). Attempts to
(Abidi and Warner, 2001) from which we were able to use for calibration a different than M[η] descriptor like the

Table 4 Composition and average values of monomer esters and dimer esters; no pentamers and higher mers were present
E1a E1ba E1cb E2a E2b E2c
1 day 4 days 7 days 1 day 4 days 7 days

% monomers 99.4 98.9 98.1 1.8 2.9 1.7


% dimers 0.5 1.0 1.4 91.1 88.0 89.3
% trimers 0.1 0.1 0.3 6.8 8.7 8.6
% tetramers 0.0 0.0 0.2 0.4 0.3 0.4
Mn (GPC1) 395 396 398 682 680 686
Mw (GPC1) 396 398 401 692 694 698
Ð (GPC1) 1.003 1.004 1.009 1.015 1.020 1.017
EN (GPC1) 0.004 0.007 0.013 1.022 1.014 1.036
Mn (GPC2) 374 380 385 648 649 655
Mw (GPC2) 379 382 409 663 663 673
Ð (GPC2) 1.013 1.005 1.062 1.022 1.022 1.028
EN (GPC2) −0.070 −0.050 −0.032 0.902 0.903 0.926
a
There was a lactone peak with area of 5.1% of the oligomer content.
b
There was a lactone peak with area of 9.5% of the oligomer content (the lactone amounts are not included in the total).

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

radius of gyration (Chance et al., 1995; Sun et al., 2004) or and


retention radius (Boyd et al., 1996) were only partially suc-
cessful. In light of these findings, the determined limiting RSGPC2=GC ¼ 3:0 ð24Þ
viscosity number values in this manuscript should be
regarded as apparent values, and a more direct method for The presence of a lactone peak (a low MW peak) leads
measuring (extrapolation from viscosity of solutions in to lowering of Mn, calculated by GPC, which leads to a
THF) should be used to confirm them. decrease in ENGPC. The deviations from one of the propor-
tionality constants in Eqs 21 and 23 could be due to the
presence of cyclic estolides. If higher cyclic estolides are
Comparison of the Methods for Determination of present, they will lead to an increase of ENGC and most
the EN probably, similar to lactones, will decrease the ENGPC
values. Calculations were made to determine the effect of
Fig. 6 compares the EN values, obtained by the the cyclic estolides on ENGC, assuming that the unidenti-
transesterification-GC and GPC1. It can be seen that there fied peaks in Eb and Ec are cyclic dimer estolides and
is a good correlation between the values. Because both Eq. 5 is valid for j ≥ 2. The result was that the cyclic esto-
methods for determining EN yield some errors in the lides will increase the ENGC values by 3.7–4.3%, which are
values, the method of Deming was used to calculate the close to the deviations of the proportionality constant from
regression coefficients (Billo, 2001). The result was: 1 (5.6% and 6.7%).
ENGPC1 ¼ 0:944 × ENGC + 0:0188 ð21Þ Therefore, the presence of cyclic estolides and lactones,
if not accounted for, will lead to opposing errors in determi-
nation of the EN.
From the slope of the regression line and standard devia- The lactones can be identified in the transesterification-
tions of the EN, determined by the two methods (σ GC and GC analysis, because they result in FAME with hydroxyl
σ GPC1), the relative sensitivity (RS) of the two methods was groups at the 4- and 5-positions, but we do not expect the
calculated according to Skiera et al. (2012): higher cyclic estolides to yield unique hydroxyl FAME
σ GC upon transesterification. In GPC2, the summations used in
RSGPC1=GC ¼ 0:944 ¼ 3:5 ð22Þ
σ GPC1 the calculation can exclude the low z interval, where the
lactone peak is observed. While the minimum between the
Because the RS > 1, it can be concluded that the GPC1
lactone peak and the monomer ester was at z ~0.954, for
method is more sensitive than the transesterification-GC for
the monomer acid the peak was ~1.0, so z = 0.900 looked
the determination of the EN.
like a better cutoff. In the current study, lactone was dis-
Comparing the GPC2 results with the transesterification-
tilled off and the monomer acid was in a low amount in the
GC (data not shown) resulted in:
investigated estolide acids, so the value of the low-cutoff
ENGPC2 ¼ 0:933 × ENGC −0:0443 ð23Þ limit had a negligible impact. In the future, the best cutoff
value for z may need to be addressed if the undistilled esto-
lide acids are analyzed.
Acids, monomer removed (A-1 a,b,c)
2.5 Un-distilled Esters (E a,b,c) In transesterification-GC, the hydroxy FAME with
Esters, monomer removed (E-1 a,b,c)
Esters, dimer reduced (E-2 a,b,c) hydroxy groups at the 4- or 5-position elute separately
2.0 Dimers(E2 a,b,c); Monomers(E1 a,b,c) and can be excluded from the calculations, too. Exclud-
EN from GPC1

1.5 ing all of the 4- and 5-hydroxy FAME in calculations


ignores estolides with these FA incorporated into the
1.0
Deming fit: estolides. A more elaborate study and reasoning needs to
of GPC1;
0.5 of GPC2. be applied to the analysis of the transesterification-GC
data in the cases where both lactones are present and
0.0
there are 4- and 5-hydroxy FA incorporated in the esto-
0.0 0.5 1.0 1.5 2.0 2.5
EN from GC lide oligomers. We are not aware of studies on such
cases.
In the case of overloading the column, the GPC1 method
Fig. 6 Comparison of the estolide numbers (EN), determined from
GC (x-axis), and the ones determined from GPC1 (y-axis). The error needs an additional adjustment parameter to account for the
bars show the standard deviations. We see that the error bars from the shift of the overloaded peak. In GPC2, there is no simple
GC are bigger than from GPC1. The symbols from the same sets from way to adjust for the overloaded peaks so the calculated
left to right correspond to a, b, and c samples. The best fit curve for
average MW and the EN are underestimates of the actual
the EN data for GPC2 is also shown (to avoid crowding the figure,
the individual datapoints from GPC2 are omitted) values.

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

A source of error in the GPC2 stems from the band


broadening of the GPC peaks. The band broadening leads p
EN ¼ ð27Þ
to a decrease of the apparent Mn of the studied oligomers 1 −p
(Busnel et al., 2001). A proportional fit between MnGPC2 Ð can be calculated from Eq. 19. The theoretical curve
and MnGPC1 (of all the datapoints, excluding the monomer from these formulas for the case of estolide esters is
and dimer) showed that GPC2 underestimates Mn by 2.6% plotted in Fig. 7. It can be seen that the experimental Ðs
compared to GPC1. The goodness of fit was R2 = 0.9996. for the undistilled estolide esters are higher than the the-
oretical ones. This cannot be easily explained, because
an inadvertent distillation of some of the monomers dur-
Mw and Dispersity ing handling and esterification would lead to lower Ð
and increased EN values for the remaining estolide
In addition to Mn (and the related EN), the GPC gives esters. It can be seen that this was the case for the dis-
information about the width of the distribution. The most tilled esters.
common method to express the distribution width is
through the dispersity Ð (Eq. 19). Fig. 7 displays the calcu-
lated Ð values as a function of EN. It can be seen that the
Dimer Esters, and the undistilled estolide esters, have close Conclusions
EN values and it is hard to distinguish them by the EN
alone, but the Ð clearly differentiates the two samples. In the current study, we demonstrated the feasibility of
Comparing the EN and Ð results from the same type of using GPC to characterize estolide acids and estolide esters.
mixture, but at different reaction times, shows that the EN An important assumption for the GPC analysis was that the
and Ð grow with the synthesis time. The largest changes MHE is valid for both the estolide acids and estolide esters.
are seen during the first 24 h, while further reaction time The good fits support the validity of this assumption. In the
leads to less changes overall. In the case of the estolide current study, we determined their MHE parameters in
esters, the differences in EN and Ð between 96 and 168 h THF at 40  C.
were undetectable. A detectable difference was seen GPC yields EN for the samples that correlated well
between Eb and Ec in the amount of lactone formed, but it with the EN, determined by the traditional method of
is not included in these calculations. methanolysis and GC (transesterification-GC). Based on
Flory’s distribution (Eq. 1) and Eqs (16, 17, and 20) can the repeatability of the values, Deming analysis picked
be used to derive the following formulas for the aver- the GPC as a more sensitive method than the traditional
age MW: one used for EN determination. Both methods do not
ΔM account for the possible presence of cyclic estolides. The
Mn ¼ + M0 ð25Þ presence of cyclic estolides will influence the calculated
1 −p
" # EN in the opposite direction, so the closeness of the
 
1+p pð1 −pÞ results from transesterification-GC and GPC indicates
Mw ¼ ΔM + M0 1 − ð26Þ that the amount of cyclic estolides was low in this study.
1 −p M0
1 + ΔM ð1 −pÞ
In addition to the EN, GPC analysis yields information
Eb&Ec about the distribution of the oligomers and can identify
1.4
Ea c the presence of a side product—lactone. Based on this,
a b
1.3 we recommend using the GPC in cases when more in-
Ð from GPC1

b&c
b&c depth information about the estolides is needed, and
a
1.2 a using both GPC and transesterification-GC when the
Acids, monomer removed (A-1 a,b,c) presence of cyclic estolides is suspected. Care must be
1.1 Un-distilled Esters (E a,b,c)
Esters, monomer removed (E-1 a,b,c) taken not to overload the GPC columns during the
Esters, dimer reduced (E-2 a,b,c)
1.0 Dimers(E2 a,b,c); Monomers(E1 a,b,c)
analysis.

0.0 0.5 1.0 1.5 2.0 2.5


EN from GPC1 Acknowledgments We thank Daniel Knetzer for help with the GPC
instrument and Linda Manthey for the help with the runs. This work
is a part of the in-house research of the Agricultural Research Service
Fig. 7 The dispersity (Ð) values as a function of EN. Both are deter- of the United States Department of Agriculture.
mined using the GPC1 method. The standard deviations of the Ð
values are smaller than the marker size. The undistilled estolide esters
(Ea, Eb, and Ec) should lie on the solid line, if they follow the Flory’s Conflict of interest The authors declare that they have no conflicts
distribution of interest.

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

References chains. Journal of the American Chemical Society, 88:3209–3212.


https://doi.org/10.1021/ja00966a006
Abidi, S. L., & Warner, K. (2001) Molecular-weight distributions of Grubisic, Z., Rempp, P., & Benoit, H. (1967) A universal calibration
degradation products in selected frying oils. Journal of the Ameri- for gel permeation chromatography. Polymer Letters, 5:753–759.
can Oil Chemists’ Society, 78:763–769. https://doi.org/10.1002/pol.1967.110050903 Re-published in:
Acklam, P. J. E. R. F. P. M. (2018, March 6). Error and scaled and (1996) Journal of Polymer Science, Part B: Polymer Physics,
34:1707–1713. DOI: 10.1002/polb.1996.922
unscaled complementary error functions and their inverses—[Web
Page]. Retrieved from https://metacpan.org/source/REID/Games- Illuminati, G., & Mandolini, L. (1981) Ring closure reactions of
Go-GoPair-1.001/Erf.pm bifunctional chain molecules. Accounts of Chemical Research, 14:
95–102. https://doi.org/10.1021/ar00064a001
Anon. (2018, March 6). Scaled Complementary Error Function,
erfcx(x), computation avoiding arithmetic overflow—VBA/Excel Isbell, T. A., & Kleiman, R. (1994) Characterization of estolides pro-
[Web Page]. Retrieved from https://stackoverflow. duced from the acid-catalyzed condensation of oleic acid. Journal
com/questions/34723644/scaled-complementary-error-function- of the American Oil Chemists’ Society, 71:379–383.
erfcxx-computation-avoiding-arithmetic-o Jacobson, H., & Stockmayer, W. H. (1950) Intramolecular reaction in
Bantchev, G. B., Moser, B. R., Murray, R. E., Biresaw, G., & polycondensations. I. The theory of linear systems. The Journal of
Hughes, S. R. (2016) Synthesis and characterization of phospho- Chemical Physics, 18:1600–1606. https://doi.org/10.1063/1.
nates from methyl linoleate and vegetable oils. Journal of the 1747547
American Oil Chemists’ Society, 93:1671–1682. https://doi.org/10. Kalambet, Y., Kozmin, Y., Mikhailova, K., Nagaev, I., &
1007/s11746-016-2909-3 Tikhonov, P. (2011) Reconstruction of chromatographic peaks
Billo, E. J. (2001) Excel for chemists a comprehensive guide (2nd using the exponentially modified Gaussian function. Journal of
ed.). New York, NY: Wiley-VCH. Chemometrics, 25:352–356. https://doi.org/10.1002/cem.1343
Boyd, R. H., Chance, E. R., & Ver Strate, G. (1996) Effective dimen- Kissin, Y. V. (1995) Molecular weight distributions of linear poly-
sions of oligomers in size exclusion chromatography. A molecular mers: Detailed analysis from GPC data. Journal of Polymer Sci-
dynamics simulation study. Macromolecules, 29:1182–1190. ence: Part A: Polymer Chemistry, 33:227–237. https://doi.org/10.
https://doi.org/10.1021/ma9508663 1002/pola.1995.080330205
Bredsguard, B., Thompson, T., Cermak, S. C., & Isbell, T. A. (2016) Milic, J., Teraoka, I., & Petrovic, Z. S. (2012) Solution properties
Estolides: Bioderived synthetic base oil. In B. K. Sharma & of biobased hyperbranched polyols investigated by multiple-
G. Biresaw (Eds.), Environmentally friendly and biobased lubri- detection size exclusion chromatography. Journal of Applied
cants (pp. 35–50). Boca Raton, FL: CRC Press. Polymer Science, 125:E586–E594. https://doi.org/10.1002/app.
36800
Busnel, J. P., Foucault, F., Denis, L., Lee, W., & Chang, T. (2001)
Investigation and interpretation of band broadening in size exclu- Noor, M.A.M.,Sendijarevic, V.,Hoong, S.S., Sendijarevic, I.,
sion chromatography. Journal of Chromatography A, 930:61–71. Ismail, T.N.M.T., Hanzah, N.A., Noor, N.M., Palam, K.D.P.,
https://doi.org/10.1016/S0021-9673(01)01159-1 Ghazali, R., Hassan, H.A. (2016). Molecular weight determination
of palm olein polyols by gel permeation chromatography using
Caceci, M. S. (1989) Estimating error limits in parametric curve fit-
polyether polyols calibration. Journal of the American Oil Chem-
ting. Analytical Chemistry, 61:2324–2327. https://doi.org/10.1021/
ists’ Society, 93:721–730. https://doi.org/10.1007/s11746-
ac00195a023
016-2812-y
Cermak, S. C., Evangelista, R. L., & Kenar, J. A. (2012) Distillation
Ruddick, C. L., Hodge, P., Zhuo, Y., Beddoes, R. L., & Helliwell, M.
of natural fatty acids and their chemical derivates. In S. Zereshki
(1999) Cyclo-depolymerisation of polyundecanoate and related
(Ed.), Distillation—Advances from modeling to applications
polyesters: Characterisation of cyclic oligoundecanoates and related
(pp. 109–140). Rijeka, Croatia: InTech.
cyclic oligoesters. Journal of Materials Chemistry, 9:2399–2405.
Cermak, S. C., & Isbell, T. A. (2001) Synthesis of estolides from oleic
https://doi.org/10.1039/a903767e
and saturated fatty acids. Journal of the American Oil Chemists’
Savitzky, A., & Golay, M. J. E. (1964) Smoothing and differentiation
Society, 78:557–565.
of data by simplified least squares procedures. Analytical Chemis-
Cermak, S. C., & Isbell, T. A. (2003) Improved oxidative stability of
try, 36:1627–1639. https://doi.org/10.1021/ac60214a047
estolide esters. Industrial Crops and Products, 18:223–230. https://
Skiera, C., Steliopoulos, P., Kuballa, T., Holzgrabe, U., & Diehl, B.
doi.org/10.1016/S0926-6690(03)00062-1
(2012) 1H-NMR spectroscopy as a new tool in the assessment of
Cermak, S. C., & Isbell, T. A. (2004) Synthesis and physical proper-
the oxidative state in edible oils. Journal of the American Oil
ties of cuphea-oleic estolides and esters. Journal of the American
Chemists’ Society, 89:1383–1391. https://doi.org/10.1007/s11746-
Oil Chemists’ Society, 81:297–303.
012-2051-9
Cermak, S. C., Isbell, T. A., Bredsguard, J. W., & Thompson, T. D.
Sun, T., Chance, R. R., Graessley, W. W., & Lohse, D. J. (2004) A
(2017) Estolides: Synthesis and applications. In M. U. Ahmad
study of the separation principle in size exclusion chromatography.
(Ed.), Fatty acids: Chemistry, synthesis, and application
Macromolecules, 37:4304–4312. https://doi.org/10.1021/ma030
(pp. 431–475). San Diego, CA: Elsevier.
586k
Chance, R. R., Baniukiewicz, S. P., Mintz, D., Ver Strate, G., &
Vivó –Truyols, G., & Schoenmakers, P. J. (2006) Automatic selection
Hadjichristidis, N. (1995) Characterization of low-molecular-
of optimal Savitzky-Golay smoothing. Analytical Chemistry, 78:
weight polymers: Failure of universal calibration in size exclusion
4598–4608. https://doi.org/10.1021/ac0600196
chromatography. International Journal of Polymer Analysis
and Characterization, 1:3–34. https://doi.org/10.1080/
10236669508009704 Appendix
Di Marco, V. B., & Bombi, G. G. (2001) Mathematical functions for
the representation of chromatographic peaks. Journal of Chroma- Generalized Exponentially Modified Gaussian (EMG)
tography A, 931:1–30. https://doi.org/10.1016/S0021-9673(01)
01136-0
Flory, P. J., & Semlyen, J. A. (1966) Macrocyclization equilibrium Equation (11) was adequate to describe the data used in the
constants and the statistical configuration of poly(dimethylsiloxane) current manuscript

J Am Oil Chem Soc (2019)


J Am Oil Chem Soc

rffiffiffi  2  
hσ π σ z −μ s σ z −μ overflow/underflow. One possibility is to use the series
yðzÞ ¼ exp − erfc pffiffiffi − from Kalambet et al. (2011), and the other is to use the con-
jτ j 2 2τ2 τ 2 τ σ
tinued fraction presentation (Anon, 2018).
ð11Þ
Another question is to calculate the position of the peak
For the meaning of the symbols, see the explanation to maximum, zp. Using that at zp, the first derivative of Eq. 11
Eq. 11 in the main text. However, as it is pointed out is zero, it can be derived that:
(Di Marco and Bombi, 2001; Kalambet et al., 2011), no rffiffiffi
jτ j 2
single formula for EMG can be evaluated for all the combi- erfcxðuÞ ¼ when z ¼ zp ðA4Þ
nations of (z − μ), σ, and τ due to numerical overflow or σ π
underflow of the functions. Kalambet et al. (2011) pre- or:
sented an algorithm to evaluate the EMG for any combina-   qffiffiffiffiffiffiffi
σ pffiffiffi jτ j 2
tion of (z − μ), σ, and τ. Here, we expand their approach to zp ¼ μ + σ −s 2erfcxinv =π ðA5Þ
the generalized EMG. First, a variable u is calculated: τ σ
s σ z −μ where erfcxinv() is the inverse function of erfcx(). For
u ¼ pffiffiffi − ðA1Þ
2 τ σ |τ|/σ > 0.001, erfcxinv() can be evaluated numerically using
an iterative procedure adapted from Acklam (2018) to
If τ is equal to zero, then just a Gaussian is calculated Visual Basic for Excel. Since Kalambet et al. (2011)
from z − μ to σ. warned that this iterative procedure cannot converge for
For u < 0, the Eq. 11 is used. The erfc() function is avail- very small arguments, a crude approximation was made to
able in the Excel and other software programs (Appendix ensure good starting values for the iterations. The Eq. A5 is
S1, Supporting information). For u > 0, the Eq. (11) can be that it is not numerically stable for very small values of
rearranged into: |τ|/σ. This can be easily overcome by using the approximate
rffiffiffi  z −μ2  
hσ π Eq. A6 (Kalambet et al., 2011):
yðzÞ ¼ exp −0:5 exp u2 erfcðuÞ ðA2Þ
jτ j 2 σ zp ¼ μ + τ for jτj=σ < 0:001 ðA6Þ
The scaled complementary error function: The relative error of using A6 for |τ|/σ < 0.001 is less
 than 0.0001 %.
erfcxðuÞ  exp u2 erfcðuÞ ðA3Þ
An Excel file with the function needed for the calcula-
can be evaluated directly for values of u < 26.5. For tions of the generalized EMG is available upon request.
u > 26.5, a different approach is needed, due to numerical

J Am Oil Chem Soc (2019)

You might also like