Modeling Whisk(e)y, Rum, Cachaça, and Tequila Maturation
A multi-year effort has culminated in the creation of a one-of-a-kind model that simulates the spirit maturation process. This novel model can simulate barrels in three-dimensions including the staves, spirits (whiskey, whisky, rum, cachaça, and tequila so far), and air gap. Inclusion of evaporation and esterification allow the prediction of spirit maturation over years in a matter of minutes on a desktop computer.
Please check out this short 7-minute video:
Barrels on different floors in a rickhouse can be simulated to understand how the proof changes along with the flavor profile.
Work by Mejia Diaz et al. provides GC-MS data for rested (3 mo.), aged (12 mo.), and very aged (36 mo.) tequila. After assuming an end date for ageing and grabbing weather data from Guadalajara, Mexico, I enhanced my chemistry model to account for leaching, diffusive, and kinetic influences that can occur during maturation. A bit of optimization of the chemistry constants and my model can predict the concentrations of the analytes within the ranges presented.
Speaking of tequila maturation...did I mention I can also simulate horizontal (side-up) vs. vertical (head-up) barrels? As the industry considers how to age their spirits, understanding the difference in storage conditions could help with optimizing yield, proof, and even taste. Yes, these are 3-D images of my barrel model where you can see the surface of the tequila after 3 years of ageing. Interestingly, the chemistry of the head-up barrels lags behind except for vanillin. More research and investigation is needed, but this is an excellent "taste" of what my model can provide.
Since proof gallons are an important facet of the spirits industry, I enhanced my maturation model to consider entry proof. A search through the literature reveals Scotch whisky maturation data in uncharred barrels of American white oak for sinapaldehyde and syringic acid by Clyne, et al. (1993) and a follow-up effort by Withers, et al. (1995) at two different entry proof levels and two dissimilar barrel sizes. Clyne, et al. state that a lower ethanol concentration was less effective at extracting sinapaldehyde, and Withers, et al. indicate that the oxygen content within the barrel directly impacts both syringaldehyde and syringic acid but did not find a statistical impact on filling proof. A deep dive into the literature (from Haeseler and Misselhorn in 1958 to Burtron-Benitez, et al. 2023) revealed conflicting findings regarding initial proof and the resulting species. The outcome of which was the inclusion of a chemical mechanism into my model that steps through the process of wood-lignin + ethanol and subsequent oxidation until the generation of syringic acid, while also simulating the influx of oxygen into the barrel. I then pulled weather data (New Keith, Scotland) from around the time of the Clyne and Withers studies to run with my barrel model, optimizing a single set of reaction constants to the data, leading to the following findings. Not much difference is seen between the ABV results, with sinapaldehyde slightly greater at a higher proof; thus, reflecting both Clyne’s and Withers’s findings.
As discussed by Gollihue et al. (doi: 10.1038/s41598-018-34204-1), the interaction between wood and distillate plays a significant role in imparting flavor to the spirit. Thus, incorporating the penetration of spirit into the staves as part of my 3-D barrel model can provide deeper insights into the maturation process, particularly how it changes with proof and ambient conditions. Overall, the depth of spirit penetration varies along the length of the wood (see the great figure in Gollihue et al.’s article: Supplementary Figure S1d). Therefore, I focused on creating a representative model that reflects accurate trends rather than pinpointing exact penetration depths at each location in the barrel. Additionally, I aimed to capture the initial phase when the distillate is first added to the barrel, causing the stave’s moisture content to increase over time. To achieve this, I included a capillary-driven flow model in my 3-D barrel simulation. This model accounts for variations in temperature, permeability, wood porosity, viscosity, and the surface tension of the spirit. In an earlier post, I simulated the differences in Scotch whisky chemistry based on entry proof. Now, beyond predicting chemical changes, my 3-D barrel model predicts that the average spirit penetration increases from 10.33 +/- 2.19 mm to 11.62 +/- 2.47 mm when the ABV is increased from 63.4% to 67.5%. As ABV increases, the density decreases, resulting in a deeper penetration since there is less gravitational force resisting the movement. However, as surface tension decreases, there is less cohesive force, which limits how far the liquid can penetrate. At the same time, lower viscosity reduces resistance to flow and accelerates the liquid’s movement through the porous wood, leading to quicker equilibrium. Check out the first 100 days of the process in the corresponding image here. Clearly, data on this process would greatly improve model predictions!
Previously, I have modeled whiskey, whisky, rum, and tequila, so now, let’s tackle cachaça (for a detailed background on this spirit, see doi: 10.3390/foods12173325). Since the barrel plays such a crucial role in the aging process through its interaction with the spirit, expanding my 3-D barrel model to accommodate different oak woods was the next logical step. Two commonly used options for spirit maturation are American Oak (Quercus alba) and European Oak (Quercus petraea). Fortunately, Castro et al. (doi: 10.1016/j.heliyon.2020.e05586) conducted valuable research in this area, providing the necessary data to run my model and calibrate a cachaça reaction mechanism. In terms of input, American Oak has a higher density with lower porosity and permeability compared to its European counterpart. This difference affects gas exchange through the barrel walls, which is critical for spirit oxidation. American oak’s lower permeability results in slower oxygen uptake compared to European oak. This variation was incorporated into the model to predict oxygen’s effect on aging rates and oxidation-related flavor developments in cachaça. Additionally, European Oak contains higher quantities of larger molecular weight ellagitannins, while American Oak is richer in smaller, lower molecular weight compounds derived from lignin. Consequently, American Oak is often preferred for spirits like bourbon, where quicker maturation is desired; whereas, European Oak is favored for slower maturation processes and subtle flavor development, typically better suited for wines. Using this information, I performed two different simulation runs adjusting for the specific properties and characteristics of each wood. The results show that the model can accurately predict the lignin-derived vanillin and syringaldehyde content in both types of wood, demonstrating another successful expansion of my barrel model.
More information about the model is available in three White Papers:
If you, or your company is interested in collaborations, please contact me at depcik at ku dot edu or connect with me on LinkedIn.