Real time analysis of biomass fermentation using

Raman spectroscopy

Wesley Thompson1, Shannon Ewanick2

Renata Bura2, Brian Marquardt1

Applied Physics Lab1 Department of Forest Resources2

University of Washington University of Washington

March 23, 2011

Bioethanol

BiodieselBiobutanol

Biopolymers

Biochemicals

Hydrogen

BiomethanolBiogas

Bio-oil

XylitolBio-adhesives

Fractionation

Biomass Selection

n Starting biomass

¨Cellulose, hemicellulose, lignin

n Pretreated substrates

¨ Solids: Cellulose, hemicellulose, lignin

¨ Liquid: Glucose, hemicellulosic sugars, inhibitorsn Hydrolysate liquor will be our focus

Components To Be Measured

n Sampling

n Analytical characterization

¨ Sensors

¨ Instruments

n Data handling

¨ Sensor fusion

¨ Multivariate modeling

n Process modeling

n Process optimization – feedback and feed forward control

Bioprocessing Challenges

Real-time Analysis of a HydrolysateFermentation Process

¨ 900 mL hydrolysate was fermented with yeast for 8 hours

¨ Additional sugar was added to the hydrolysate

n 30 g/L glucose and 15 g/L xylose added broth (which already contained ~ 5g/L of each)

¨ The fermentation broth was held at 30°C and pH 6 for the duration of experiment.

¨ Raman spectra were collected every 1 minute

n 3/8” ballprobe with sapphire optic, 215mW power at probe tip

n Average of six, 10 second exposures

¨ HPLC Samples

n Collected a 1mL aliquot every 10 minutes during fermentation

Lignin Fluorescence Data Pretreatment

n Fluorescence is a different optical process than Raman spectroscopy¨ Can be removed while

not affecting Raman scattering information

¨ Bring all of the samples to same baseline

¨ Iteratively removed a 4th order polynomial

0.1 0.2 0.5 1 2 5 10 20 g/L

Increasing Fluorescence

400 600 800 1000 1200 1400 1600 1800

0.5

1

1.5

2

x 105

Wavenumbers/cm-1

Sig

nal In

ten

sit

y

Lignosulfonate Raw Spectra

400 600 800 1000 1200 1400 1600

0.2

0.4

0.6

0.8

1

1.2

Wavenumber/cm-1

Sig

nal In

ten

sit

y

Lignosulfonate Baselined Spectra

Fluorescent Raman Spectra

Polyfit ROI

Polynomial Removal Pretreatment

Surface ROI

Cosmic Rays

4th Order Polynomial

Fermentation Surface Plot

9001000

1100

800

Raman Shift (cm-1)

Sample

Inte

nsity (

Arb

. U

nits)

Peak 1 Peak 2 Peak 3 Peak 4

Hydrolysate Engineering Plot

0 50 100 150 200 250 300 350 400 450 5002000

2500

3000

3500

4000

4500

Sample

Inte

ns

ity (

Arb

. Un

it)

Peak 1Peak 2Peak 3Peak 4

Equilibrium3hrs 43min

Raman Ethanol Calibration

0 2 4 6 8 10 12 14-2

0

2

4

6

8

10

12

14

Measured (mg/ml)

Pre

dic

ted

(m

g/m

l)

R2 = 0.9972 Latent VariablesRMSEC = 0.22033Calibration Bias = -1.7764e-015

Raman Glucose Calibration

0 5 10 15 20 25 30 35-5

0

5

10

15

20

25

30

35

Measured (mg/mL)

Pre

dic

ted

(m

g/m

L)

R2 = 0.9992 Latent VariablesRMSEC = 0.2832Calibration Bias = -1.7764e-015

Real-time Analysis of a Simulated Batch Fermentation Process

n Simulated Fermentation liquor¨ 900 mL of synthetic sugar water was

fermented with yeast for 8 hours¨ The solution of sugar was held at

30°C and pH 6 for the duration of experiment.

¨ Glucose additions were added when ethanol peak equilibrated. The final total sugar concentration in the reactor was 25 g/L glucose

n NeSSI Fast-Loop¨ Kaiser Optical Systems Raman

n ½” ballprobe with sapphire optic, 250mW power at probe tip

n Average of six, 5 second exposures

¨ Mettler React-IRn MCT Detectorn Silver Halide Immersion Proben Diamond ATR

¨ Fiber Optic O2 Sensor¨ Cassini O2 Sensor

HPLC Concentration Plot

0 50 100 150 200 250 300 350 400 4500

2

4

6

8

10

12

Time (Min)

mg

/mL

Acetic AcidEthanolGlucoseGlycerolGlucose AdditionHPLC Sample

Raman Raw

Raman Shift (cm-1)

Inte

nsity

(A

rb.

Uni

ts)

Spectra collected before inoculationRemoved from model

Raman Polyfit

400 600 800 1000 1200 1400 1600 18000

2000

4000

6000

8000

10000

12000

Variables

Data

Raman Shift (cm-1)

Inte

nsity

(A

rb.

Uni

ts)

4th Order Polynomial

Ethanol HPLC and Raman PC1

0 50 100 150 200 250 300 350 4000

2

4

6

8

10

12

Time (Min)

mg

/mL

-8

-6

-4

-2

0

2

4

6

8

Sc

ore

Glucose HPLC and Raman PC2

0 50 100 200 250 300 350 4000

1

2

3

4

5

6

mg

/mL

150-3

-2

-1

0

1

2

3

4

Time (Min)

Sc

ore

Raman Loadings

400 600 800 1000 1200 1400 1600 1800-0.05

0

0.05

0.1

0.15

Raman Shift (cm-1)

Loa

din

g

PC1PC2

Pure Raman Spectra

400 600 800 1000 1200 1400 1600 18000

1

2

3

4

5

6

7x 104

Raman Shift (cm-1)

Inte

ns

ity

(A

rb.

Un

its

)

Ethanol

Glucose

Raman Ethanol Calibration

0 2 4 6 8 10 12-2

0

2

4

6

8

10

12

Measured Ethanol (mg/mL)

Pre

dic

ted

Eth

an

ol

(mg

/mL

)

R2 = 0.9912 Latent VariablesRMSEC = 0.31379

Raman Glucose Calibration

0 1 2 3 4 5 6-1

0

1

2

3

4

5

6

Measured Glucose (mg/mL)

Pre

dic

ted

Glu

co

se

(m

g/m

L)

R2 = 0.9832 Latent VariablesRMSEC = 0.23895

IR Spectra

Model Region

IR Scores

50 100 150 200 250 300 350 400-0.025

-0.02

-0.015

-0.01

-0.005

0

0.005

0.01

0.015

0.02

0.025

Sample

Sco

res

PC1PC2PC3

IR Loadings

10001050110011501200125013001350140014501500-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

Variable

Loa

din

gs

PC1PC2PC3

O-H Bending1440-1220 cm-1broad weak IR bandLin-Vien, 1991

IR Ethanol Calibration

0 2 4 6 8 10 120

2

4

6

8

10

12

Measured Ethanol (mg/mL)

Pre

dic

ted

Eth

an

ol

(mg

/mL

)

R2 = 0.9852 Latent VariablesRMSEC = 0.40324

IR Glucose Calibration

0 1 2 3 4 5 6-1

0

1

2

3

4

5

6

Measured Glucose (mg/mL)

Pre

dic

ted

Glu

co

se

(m

g/m

L)

R2 = 0.9472 Latent VariablesRMSEC = 0.42344

n Enzyme development¨ High consistency in solids hydrolysis¨ Analysis of conversion efficiency

n Fermentative organism development¨ Non-Destructive monitoring of organisms¨ Rapid screening of organisms for pentose fermentation

n Existing bioethanol plants¨ Over 130 in US ¨ Require means to monitor glucose and ethanol levels in real time

during hydrolysis and fermentation

Potential Applications

NeSSI Fast-loop Design

NeSSI Sterilizationn Began by autoclaving everything

¨ NeSSI top mounts were removed from substrate

¨ All fittings, screws and components autoclaved

¨ Tubing autoclaved

n Removed from autoclave and immediately put into sterile hood

¨ 15 minute UV lamp cycle

¨ Ethanol spray all surfaces

n Flowed concentrated yeast solution through block with peristaltic pump for 10 min

n Rinsed with 500 mL of 10% bleach solution to waste

n Flowed bleach solution for 10 min

n Closed valves and let bleach sit in block and tubing for 1 hour

n Rinsed with 1 liter of 0.2 micron filtered sterile water

n Rinsed with 1 liter of sterile media solution (autoclave)

n Filled with sterile media solution

n Incubated with valves closed and all ports capped with high pressure fittings for 3 days at 37 °C

n Streaked plates from all top mount and substrates

Rinsed Filled

Incubated Swabbed

Culture plates

1. End piece (Start Flow)2. Valve substrate3. Top mount 14. Substrate 15. Top mount 2

6. Substrate 27. Top mount 38. Substrate 39. Valve substrate10.End piece (End Flow)

1 2

3 5 7

94 6 8 10

contaminated

Sterile Media ControlStreaked before filling NeSSI block, no colonies

Streaked after filling NeSSI blockNote: colonies not visually similar to colonies in NeSSI block. External contamination?

Conclusions and Challenges

n It is possible to sterilize NeSSI components

¨ The method is not perfect yet

n Possibly contaminated by residual cells trapped in valve seat

n Pre autoclaving worked to remove majority of possible contamination points

¨O-rings and tubing

n Must maintain better sterile practice on media as control

Applied Physics Lab

CPAC

Charlie Branham, Lauren Hughs, Michael Roberto

Kaiser Optical Systems

Ian Lewis

UW Forestry

Rick Gustafson

Renata Bura

Shannon Ewanick

Acknowledgements