1

Role of macropore flow in the transport of 1

Escherichia coli cells in undisturbed cores of a 2

brown leached soil 3

4

Jean M.F. Martins*, Samer Majdalani#, Elsa Vitorge, Aurélien Desaunay, Aline 5

Navel, Véronique Guiné, Jean François Daïan, Erwann Vince, Hervé Denis and Jean 6

Paul Gaudet 7

8

Laboratoire d'Etudes des Transferts en Hydrologie et Environnement (LTHE, UMR 5564) – 9

CNRS-INSU / Univ. Grenoble I / INPG / IRD – Domaine Universitaire BP53 - 38041 Grenoble 10

Cedex 9, France 11

12

* Corresponding author: jean.martins@ujf-grenoble.fr, Tel.: +33476635604 13

14

# Present address: HydroSciences Montpellier, UMR 5569, Université Montpellier 2, CC MSE, 15

34095 Montpellier Cedex 5, France. 16

17

18

19

20

Keywords: 21

Bacteria transport, undisturbed soil cores, breakthrough curves, porosity structure, mercury 22

intrusion porosimetry, repeated injections23

2

Abstract 1

The objective of this work was to evaluate the transport of Escherichia coli cells in undisturbed 2

cores of a brown leached soil collected at La Côte St André (France). Two undisturbed soil cores 3

subject to repeated injections of bacteria cells and/or bromide tracer were used to investigate the 4

effect of soil hydrodynamics and ionic strength on cells mobility. Under the tested experimental 5

conditions, E. coli cells were shown to be transported at the water velocity (retardation factor close 6

to 1) and their retention appeared almost insensitive to water flow and ionic strength variations, 7

both factors being known to control bacteria transport in model saturated porous media. In contrast, 8

E. coli breakthrough curves evolved significantly along with the repetition of the cells injections in 9

each soil core, with a progressive acceleration of their transport. The evolution of E. coli cells 10

BTCs was shown to be due to the evolution of the structure of soil hydraulic pathways caused by 11

the repeated water infiltrations and drainage as may occur in the field. This evolution was 12

demonstrated through mercury intrusion porosimetry (MIP) performed on soil aggregates before 13

and after the repeated infiltrations of bacteria. MIP revealed a progressive and important reduction 14

of the soil aggregates porosity, n, that decreased from approximately 0.5 to 0.3, inducing a 15

decrease of the soil percolating step from 27 to 2 µm. This result evidenced a clear compaction of 16

soil aggregates that concerned preferentially the pores larger than 2 µm equivalent diameter, i.e. 17

those allowing bacterial cells passage. Since no significant reduction of the global soil volume was 18

observed at the core scale, this aggregate compaction was accompanied by macropores formation 19

that became progressively the preferential hydraulic pathways in the soil cores, leading to 20

transiently bi-modal bacterial BTCs. The evolution of the soil pore structure induced a 21

modification of the main hydrodynamic processes, evolving from a matrix-dominant transfer of 22

water and bacteria to a macropore-dominant transfer. This work points out the importance of using 23

undisturbed natural soils to evaluate the mobility of bacteria in the field, since the evolving 24

hydrodynamic properties of soils appeared dominating most physicochemical factors. 25

26

3

1. Introduction 1

Bacterial cells are known to be mobile in the subsurface enabling colonization processes or 2

dispersion of pathogenic cells. 1-5 The dynamics of microorganisms in the unsaturated zone of soils 3

can also play significant roles in geochemically important processes such as the transport and 4

removal of trace elements and anthropogenic contaminants. 4-8 Bacteria or viruses may present 5

themselves a risk for the environment and especially for aquifers 5,9-15 Most waterborne pathogens 6

such as viruses or bacteria are of faecal origin, and come from human and/or animal wastes. The 7

quality of groundwater subtracted from drilled wells as both drinking and irrigation water resource 8

can be affected by the movement of pathogenic bacteria if water extraction wells are located close 9

to soil surfaces subject to microorganisms’ land application through pit latrines, cattle grazing, 10

sewage sludge spreading, or in situ bio-augmentation in contaminated sites. 16,17 This increasing 11

release of microorganisms into the environment by human applications of solid wastes (e.g. 12

manure) and contaminated wastewater effluents has led in some cases to calamities that 13

demonstrate unexpected breakthrough of harmful microbes thus imposing the necessity to better 14

understand and predict pathogenic microorganisms transport through the vadose zone. 10-12,16,17 15

Microbial cells transport in porous media has received important attention in the last decades, 16

especially through laboratory approaches that have evidenced effects of factors such as solution 17

chemistry, flow velocities or cell size or morphology, hydrophobicity or surface charge. 5,7,14,18-20 18

However, there is still a lack of knowledge for understanding the processes that control in situ the 19

survival and reactive transfer of biotic colloids in the unsaturated zone of soils, which are critical 20

points for the persistence of a biological risk. 14,19,21. Indeed, the risk associated with pathogenic 21

bacteria contaminations is controlled by both the natural filtration capacity of soils (irreversible 22

retention and mobility) but also by the relative inability of most enteric bacteria to survive longer 23

than a few days or weeks in natural soils. 1,21 The modification of one of these two properties may 24

impact strongly the risk of groundwater pollution with pathogenic bacteria. A good indicator of 25

pathogens in groundwater is the bacterium Escherichia coli that is also easy to detect. 22,23 Due to 26

4

their colloidal size and the negative surface charge of their cell wall, E. coli cells are mobile and 1

able to travel long distances in the subsurface. 24,25 This is the reason why these bacteria are the 2

most widely used for evaluating bacterial transport in porous media and for testing mathematical 3

models of microbial transport that are mostly based on either the classical colloid filtration theory 4

or Langmuirian blocking processes or geometrical retention processes such as straining. 10,14,15,24,26-5

31 Factors affecting (bio)colloid transport (e.g., surface charge, ionic strength and solution 6

chemistry, pore size exclusion, Darcy flux or median of the grain size distribution) have been 7

largely documented. 5,10,32-36 Most of these studies on bacteria transport were conducted with 8

homogeneous porous media (model sand, glass beads...) under saturated conditions. 28,30,31 9

Knowledge acquainted with model media like pure sand matrices, cannot be transferred easily to 10

natural porous media. Difficulties arise in natural soils due to the anisotropy and the heterogeneity 11

of these porous media but also because of the temporal evolution of their properties such as their 12

spatial structure (micro and macro-porosity) in relation with field wetting/drying cycles or clay 13

swelling that can strongly modify the hydraulic pathways controlling water movement. In the field, 14

unsaturated soil heterogeneity and macropore water transport are suspected to play a crucial role in 15

microbial cell mobility. 37-40 This may explain the frequent discrepancies observed between field 16

measurements and bacteria transfer predictions. Namely, size exclusion effects may be drastically 17

more pronounced at large observation scales in natural media, 19,41 where bacteria and colloids in 18

general can be subject to matrix, macropore, by-pass and non-equilibrium flow and move with the 19

pore-water in both the macro and micro-porosity of soils. Understanding the importance of 20

macropore transport of bacteria in natural soils in relation with the preferential flow degree, will 21

help improving microbial transport models and better predicting aquifer vulnerability to bacterial 22

contamination. 12,13,19,26,37-43 23

The aim of the present work was to investigate the transport of pathogenic cells in natural soils 24

using nontoxigenic E. coli cells as surrogate to the pathogenic bacterium. Repeated applications of 25

5

E. coli cells were performed on two undisturbed soil cores under transient conditions simulating 1

repeated land effluent applications. The effect of boundary and initial conditions (evolution of 2

pore-network in relation with repeated infiltration-drainage cycles) as well as that of the solution 3

ionic strength and water velocity on the mobility of E. coli cells were evaluated through column 4

transport experiments. Soil and bacterial cells properties were also characterized in order to model 5

their potential interactions with soil constituents. The importance of characterizing soil pore size 6

distribution that controls dominant hydraulic pathways (in micro or macropores) in undisturbed 7

natural soils is particularly discussed. 8

9

2. Material and methods 10

2.1 Bacteria 11

Escherichia coli strain K12 MG1655 was used as surrogate to pathogenic enteric bacteria. It 12

contains the plasmid pGFPuv (Ozyme) making the cells fluorescent. The bacteria were grown 13

overnight at 37°C in liquid Luria Bertani medium supplemented with ampicillin (100µg mL-1) for 14

cell selection. After growth, the cells were harvested by centrifugation at 9000g and rinsed twice in 15

deionised water (DIW). The bacterial cells were then resuspended in DIW or KBr solution at 8 10-16

3 or 10-1 M in order to evaluate the effect of soil solution ionic strength on bacteria transport. The 17

injected cell concentration was set to an optical density of 1 (corresponding to 109 cells mL-1) at 18

600 nm, 44 as measured with a UV-VIS spectrophotometer (Lambda 25 Perkin Elmer). Before 19

bacteria injection, the cells were inactivated by addition of chloramphenicol (2 µg ml-1) in the cell 20

suspension for one hour. This procedure inactivates the protein synthesis thus avoiding cell 21

proliferation during the breakthrough experiments due to the presence of natural organic carbon in 22

soil solution that promotes bacterial growth. Inactivating the cells permits the total E. coli cells 23

number remaining constant during the breakthrough experiments and renders possible the 24

calculation of cell mass balances. 25

6

The size distribution and Zeta potential (ζ) properties of the model Escherichia coli cells were 1

determined with a laser granulometer Mastersizer 2000 (Malvern Instruments) and a Nano ZS 2

zetameter (Malvern Instruments), respectively. These measurements were performed under the 3

different geochemical conditions used in the experiments and are described in detail in the 4

Supporting Information section (SI). 5

Bacteria enumeration was performed following the procedure described by Martins et al., 45 6

modified by Baptist et al., 46 using an epifluorescent microscope (Axioscope, Zeiss France) 7

equipped with a digital camera (DP50, Zeiss France) for image recording. The columns effluents 8

and bacterial suspensions were diluted in sterile NaCl 0.9 %. 1mL of the cells suspensions was 9

filtered on 0.2 µm polycarbonate black membrane filters (Millipore). Bacteria were enumerated at 10

1000-fold magnification by direct counting under UV excitation at 488 nm. With this protocol, the 11

precision of the enumeration is about ± 10 %. 12

13

2.2 Soil 14

The studied soil is a brown leached soil under culture of maize {corn ?} collected at the 15

experimental field site of La Côte-Saint-André (F) (45°23’N, 5°15’E). The main hydrodynamical 16

properties of the LCSA soil have been already described 47 whereas its physical chemical 17

properties are presented in Table 1. To measure the soil Zeta potential, a sample of soil was 18

vigorously shaken in the chosen electrolyte and an aliquot of the supernatant was collected 19

(decanted from the vial) and analysed by zetametry (Zetasizer 3000HS, Malvern Instruments). Two 20

undisturbed soil cores, named A and B, were collected at the field site by pushing in the soil PVC 21

tubes (15 cm in diameter, 25 cm in length). The two cores were covered at both sides to limit 22

evaporation, and carefully transported to the laboratory where the experimental bacterial 23

breakthrough procedure started immediately at 23°C (The soil initial water content was 0.22). 24

The pore size distribution of the LCSA soil inside the cores was measured by Mercury Intrusion 25

Porosimetry (MIP), a well established technique used for years to provide pore diameters 26

7

distribution in porous media. Mercury intrusion tests were performed on soil aggregates (triplicate) 1

of about 1 g in mass, oven-dried during one week at 50 °C to empty the pores avoiding shrinking. 2

The porosity measurements were performed with a Micromeritics Autopore IV porosimeter in a 3

range of pressures varying from 0.0035 to 200 MPa corresponding, for cylindrical pores, to 4

diameters between 350 and 0.0063 µm. We used the XDQ multi-scale network model 5

(http://www.lthe.fr/LTHE/spip.php?article226), based on percolation theory, to represent as close 6

as possible the soil pore network. 7

8

2.3 Experimental setup 9

The experimental setup used for the study of E. coli cells breakthrough was adapted from 10

Majdalani et al., 48 and is presented in Figure 1. It consists of the undisturbed soil cores supplied at 11

the top with aqueous solutions of variable ionic strength through a rain simulator constituted of 8 12

hypodermics needles (aperture of 0.5 mm). The needles were distributed on a square network and 13

placed about 3 cm above the soil surface. A peristaltic pump controlled rainfall duration and 14

intensity: 10 or 20 mm h-1 (corresponding to heavy rain conditions, considered as a worst case that 15

may occur immediately after a field application of bacteria). The bottom of the column was 16

positioned on a perforated stainless-steel grid of 2 mm mesh, onto which a funnel was sealed and 17

connected to a fraction collector (ISCO) that permitted the collection of 200 mL fractions of 18

effluent solution. Between two rainfall events, soil surface was covered while drainage was let free 19

at the bottom of the column between two consecutive infiltrations, allowing the soil to drain from 20

0.46 (ƟS) to 0.28 (ƟFC), the field water holding capacity of the soil. 21

2.4 Experimental procedure 22

All the breakthrough experiments (Table 2) consisted in the injection of 800 mL (45 mm height) of 23

a suspension of chloramphenicol-inactivated E. coli cells into the soil cores under constant rainfall 24

intensity (steady state hydrodynamic regime) of 10 (4.5h) or 20 mm h-1 (2.25h). During the pulse 25

injection of E. coli cells, the suspension was agitated at 200 rpm in order to prevent cell 26

8

sedimentation and to guarantee the homogeneity of the injected cells suspension all along the 1

injection duration. Then, the soil column was flushed with sterile DIW for 40 or 20 hours, at 10 or 2

20 mm h-1, respectively (injection of 400 mm of water). Bacterial injections were repeated 6 and 4 3

times on column A and B, respectively, under identical experimental conditions (i.e. same 4

incoming solution and rainfall intensity) to evaluate the behaviour of bacteria when injected 5

repeatedly in the two undisturbed soil cores (Table 2). Experiments were separated by a two days 6

interval, during which the soil surface was covered and drainage was let free at the bottom of the 7

column. In column B, after 4 identical bacteria injections, the salinity of the bacterial suspension 8

and rainfall intensity were varied in order to study the effect of these two factors on E. coli cells 9

breakthrough in soils. For this we performed three bacterial breakthrough experiments with 10

incoming solutions of increasing ionic strength (10-7, 8 10-3 and 10-1 M KBr). The use of bromide 11

ions permitted simultaneously to trace water movement and to characterize soil hydrodynamics by 12

model fitting. A last experiment was conducted at 10 mm h-1 with 8 10-3 M KBr incoming solution 13

to evaluate the effect of the water flow on the transport of bacterial cells in undisturbed soil 14

columns (Table 2). Bacterial breakthrough curves (BTC) were constructed by measuring cell 15

concentrations in the effluents and by weighing the eluted soil solution. BTC are presented in a 16

dimensionless form (C/Co and V/Vo) by dividing the outlet concentration by the inlet 17

concentration and the eluted volume by the pore water volume, Vo, respectively. Vo was 18

determined gravimetrically by weighing the soil cores at the end of each infiltration experiment 19

(water saturated) and after full drying of the core at 105°C at the end of the study. 20

2.5 Chemical analysis. 21

Water tracer experiments were conducted with the bromide ion (KBr) applied to the columns at 8 22

10-3 M following the procedure described by Martins and Mermoud. 49 Bromide concentrations 23

were determined by Capillary Electrophoresis (Waters). With this procedure, the detection limit of 24

the bromide analysis was 100 ng/L. 25

26

9

3. Results and Discussion 1

3.1 Characterization of the bacterial cells properties 2

3.1.1 Cells size distribution 3

Results on the size distributions (equivalent spherical diameter) of E. coli cells, measured by 4

Dynamic Light Scattering (DLS), are presented in Figure S1 (SI). The coefficient of uniformity of 5

the cells was 1.96, indicating a quite high uniformity of the size of the injected cells. The d50 6

parameter of the cells determined by DLS was 1.14 µm +/- 0.02, whatever the ionic strength tested 7

(DIW and 8.10-3M or 0.1 M KBr), indicating that no significant cell aggregation occurred, even at 8

the highest ionic strength. This result was confirmed by TEM observations and image analysis 9

(data not shown) that revealed a majority of isolated cells and permitted to determine the average 10

lengths of the major and minor axes of E. coli cells: length = 2.1 µm ±0.2 µm and width = 0.6 µm 11

±0.1 µm. 12

13

3.1.2 Cells Zeta potential 14

Figure S2 (SI) and Table 3 present the Zeta potential, ζ, (related to electrophoretic mobility) of the 15

cells and of the LCSA soil measured in DIW and in 8.10-3 M and 0.1 M KBr solutions. E. coli cells 16

appeared strongly negatively charged in DIW (-57 mV) and remained negatively charged even at 17

0.1 M KBr (-6 mV). The smallest constituents of the LCSA soil appeared also negatively charged 18

(-20.1 to -6 mV), whatever the ionic strength tested. These results are in agreement with a large 19

number of studies that have shown that almost all living cells and particles of non volcanic soils 20

are negatively charged at “environmental” pH values. 12,24,50,51. 21

The measured Zeta potentials of the cells and the soil constituents permitted to calculate DLVO 22

interaction parameters (Table 3) and interaction energy profiles (Figure S3, SI) at the 3 ionic 23

strengths tested using the approach proposed by Redman et al. 35 The Hamaker constant was set to 24

1.20 10-20 J. The results showed that in both DIW and 8 10-3 M KBr solution, energy barriers exist 25

(Φmax), which limit interactions of the cells with the soil constituents. However, in both conditions 26

10

a secondary minimum of energy of variable height (Φmax) and depth (Φmin2) exists that may retain 1

bacterial cells in the soil, especially at the intermediate ionic strength. At 0.1 mM KBr, no energy 2

barrier exists indicating the potential occurrence of strong interactions between cells and soil 3

constituents. 4

5

3.1.3 Cells fluorescence stability 6

Before performing the bacteria breakthrough experiments we evaluated the stability of the 7

fluorescence of the inactivated cells provided by the gfpUV plasmid, since it may vary during the 8

experiments and thus affect the validity of the bacterial enumeration results. For that, we 9

enumerated fluorescent E. coli cells over time in a water suspension of chloramphenicol-10

inactivated bacteria kept at room temperature (data not shown). The measured number of 11

fluorescent cells remained constant for a period of over 12 days, the total number of injected 12

bacteria neither increased (cells were inactive) nor decayed (stability of the fluorescence labelling), 13

and then decreased rapidly after 15 days. Consequently, the duration of breakthrough experiments 14

was set to 24 to 48 hours to get confident numbers of bacterial cells and mass balance results. 15

16

3.2 Bacterial transport in undisturbed soil cores 17

The mobility of E. coli cells in the undisturbed soil columns of the LCSA soil was evaluated by 18

performing 6 (column A) or 4 (column B) repeated pulse injections of 800 mL of bacteria 19

suspensions at a concentration of 109 inactivated cells mL-1. The breakthrough curves of bacteria 20

obtained after the successive cells injections in each independent soil core are presented in Figure 21

2. The results show that the calculated mass balance of the bacterial cells varied between 90 and 22

109 % for all successive injections, showing that almost all injected E. coli cells were recovered in 23

the effluent of both undisturbed soil cores (within the enumeration error of ± 10 %). Figure 2 24

shows that the shape of bacteria BTC was not constant and evolved progressively all along the 25

successive bacteria injections in a similar way for the two independent cores, indicating that the 26

11

observed behaviour is reproducible. The first E. coli BTC in both columns A#1 and B#1 presented 1

a quite classic and regular shape, with mono-modal curves significantly dispersed. Both BTC 2

presented a tailing, indicative of slow restitution of bacterial cells previously retained in low flow 3

zones or onto solid surfaces (probably in secondary minimum energy sites). The bacterial cells 4

begin to leach out of the soil columns very fast after their injections (at 0.3 pore volumes). The 5

peak was positioned at 0.5 pore volumes, indicating that the bacterial cells were not transported 6

through the whole volume of soil pore water. 7

During the second infiltration (#2), the two BTC presented a bi-modal shape, with a small peak 8

appearing in the rising front of the curves A#2 and B#2, indicating a partial fast transport of 9

bacterial cells combined with a still dominant slow transport of the cells in the soil matrix. This 10

dual mode of transport of bacteria (fast and slow) was assumed to relate to a progressive 11

modification of the soil structure and to the settlement of preferential flow pathways in the two 12

undisturbed soil cores. This preferential flow of bacteria strongly increased at the third application 13

of bacteria (BTC #3) until it became the dominant transport process of the E. coli cells in BTC #5 14

and #4 of cores A and B, respectively. The acceleration of bacteria transport from one infiltration 15

event to another (Figure 3) was supposed to reflect the temporal evolution of the hydrodynamic 16

active pore network within the soil. This evolution affected the mean residence time of the 17

bacterial cells in the cores, which decreased from 4.7h +/-0.1 to 2.5h +/- 0.5 as determined from the 18

first order moment of the BTCs. 19

20

3.3 Evolution of the LCSA soil properties 21

To validate the assumption of an evolution of the hydrodynamic properties of the soil inside the 22

cores during the successive infiltrations, we measured the particle and the pore size distribution 23

curves (PSD) of the surface soil of the two cores, before and after the repeated bacteria injections 24

(Figure 4). The results showed that both the structure and the hydraulic properties of the soil were 25

12

changed by the successive infiltrations. Indeed, Figure 4A clearly shows that soil aggregates larger 1

than 80 µm were significantly disaggregated into smallest particles by the repeated water 2

infiltrations. The evolution of the pore size distribution within the soil aggregates (Figure 4B) 3

confirms this effect as it shows a reduction of the inter-particle pore volume of the aggregates, 4

especially distributed among the largest soil pores. The soil porosity (pores thicker than 2 µm that 5

allow bacterial cells passage) was thus strongly reduced by the successive infiltrations, inducing a 6

soil compaction that is well known in soil mechanics. Since the total volume of soil inside the 7

cores didn’t vary during the breakthrough experiments, we assumed that the compaction of soil 8

pores inside soil aggregates was accompanied by the formation of macropores around the 9

aggregates, which became progressively the preferential soil hydraulic pathway, as evidenced by 10

the observed transiently bi-modal BTC of bacterial cells in experiments #2 to #4 of both cores. 11

Used the XDQ model the PSD curves presented in Figure 4B permitted to calculate theoretical 12

percolation steps of 27µm and 2µm, for the first and the last infiltration events, respectively. 13

Starting from a pore network containing a large panel of pore sizes (#1), where bacteria were 14

eluted quite slowly, at the end of the infiltrations the soils of cores A and B presented pore 15

networks dominated by macropores (A#5 and B#4). At this stage, bacteria were eluted very fast in 16

the macropores and then the curve shape returned to a mono-modal type with less tailing. This 17

shows that in the soil cores bacteria were then transported faster through preferential flow 18

pathways with residence times decreasing from 4.7h +/- 0.1 to 2.5h +/- 0.5), bypassing the water 19

flow in the soil matrix. This evolution of the bacteria transport mode passed through a transition 20

stage where bacterial preferential flow through macropores occurred simultaneously with cell 21

transport in the matrix pore network. The hydrodynamics of columns A and B evolved similarly 22

although not with the same kinetic (stabilisation after 6 or 4 infiltrations), probably because of the 23

existence of soil heterogeneity in the field. The main consequence of the soil hydrodynamics 24

evolution is a progressive acceleration of E. coli cells transport that is likely to occur in the field, 25

13

especially in natural clayey or swelling soils, which hydraulic pathways are known to evolve with 1

time. Such fast flow velocities have been already reported for bacteria or microspheres. 1,38-42 2

Altogether, the results on the evolution of particle and pore size distribution curves show that the 3

structural and, consequently, the hydrodynamic properties of the soil changed during the repeated 4

water infiltrations. Many studies have described such evolutions of soils’ pore structure at the field 5

scale and their effect on soil hydraulic conductivity and water or chemicals retention. 52-55 6

The latter soil properties have also been described as strongly controlling colloid movement in 7

unsaturated soils. 20,56 Since these two soil properties appear to evolve at the field at least at short 8

(rainfall event) and long (season or soil usage) time scales, these results tend to show that bacterial 9

or more generally colloid transport in natural unsaturated soils would evolve with the evolution of 10

soil physical properties. 11

12

3.4 Hydrodynamic behaviour 13

Before evaluating the effect of water flow and ionic strength on the transfer of bacteria in the 14

undisturbed soil cores, we characterized the hydrodynamics of the soil once stabilized. For that we 15

realized a water tracing experiment using the bromide ion (Figure 5). The bromide breakthrough 16

curves were fitted with CXFTIT 2.0, 57 considering a convection-dispersion approach with water 17

fractionation between mobile and immobile phases. 6,49,58 Figure 5 shows that after stabilisation of 18

the hydrodynamics in the soil, water (and tracer) transfer can be represented by a dominant rapid 19

flow in macropores, corresponding to the early peak of the tracer BTC and an exchange with the 20

microporosity of the soil, corresponding to the long tailing of the curve. The hydrodynamic 21

parameters of the undisturbed soil determined from the duplicate bromide tracing experiments B#5 22

and B#6 are presented in Table 4. The similarity of the parameters set confirms the stabilization of 23

the hydrodynamics of the soil in a dominant macropore flow transfer mode of water. Indeed, the 24

very low β value indicates that the effective mobile water content is low: only 14 % of the total 25

14

water content of the soil contributes to the water flow. Such a low β value confirms that water and 1

its associated bacterial cells travel preferentially in the soil macropores that contain less than 14 % 2

of the total water content. The mobile-immobile exchange coefficient α, was found equal to 0.056 3

h-1. This high α value indicates a fast exchange kinetic of bromide ions between the mobile and 4

immobile water fractions in agreement with the important tailing observed in the tracer BTCs. 5

Since bacteria BTC presented much less tailing (Figures 5 and 6), it can be concluded that because 6

of size exclusion effects, bacterial cells hardly diffuse towards the immobile water fraction of the 7

soil that is mostly present in the smallest pores of the soil matrix. This result indicates that bacterial 8

cells remain in the mobile water fraction, and explain their faster transport as compared to the 9

tracer transport. The calculated dispersion coefficient was also found high, which is conventional 10

in natural soils. 49 11

The determined soil hydrodynamic parameters confirm thus the important partitioning of soil water 12

and the low content of mobile water that is mostly located in the macropores of the soil. 13

3.5 Effect of ionic strength on bacteria transport 14

The ionic strength of bacterial suspensions varied from 10-7 M (DIW) to 10-1M (KBr) (Table 2, 15

B#4-6). The results presented in Figure 6 showed that bacterial transport is quite different from that 16

of the water tracer and confirms that in the undisturbed LCSA soil, bacterial cells are subject to 17

size exclusion and are transported mainly in the macropores (the percolating step is only of 2 µm in 18

the aggregates), whereas the tracer BTC presents a strong tailing indicating the existence of 19

significant exchange processes (α parameter) with the matrix pore water, as discussed above. 20

Figure 6 also shows that under our experimental conditions, E. coli transport seems not affected by 21

the ionic strength, as the three BTCs obtained at 10-7, 8.10-3 and 10-1M are not significantly 22

different. This result is quite surprising, especially for the latter condition for which no energy 23

barrier between cells and soil constituents was evidenced (Figure S3, SI, and Table 3). At the two 24

lowest ionic strengths, the energy barriers evidenced in Table 3, should explain the absence of cell 25

15

retention in the soil cores. The absence of effect of the highest ionic strength, which is too low to 1

induce cell aggregation as shown in Figure S1 (SI), is intriguing since several studies conducted 2

with model porous media have shown that the increase of suspension salinity can favour bacterial 3

cells aggregation as well as their interactions with the soil constituents, leading to increased cell 4

retention. e.g. 5,15,32-34 Since the effect of salinity on bacteria transport was studied once the 5

hydrodynamics of the soil cores were dominated by water flow in the macropores, our results 6

suggest that the fast transport of the cells within the soil macroporosity seems to have 7

counterbalanced the expected increased retention effects due to the high solution salinity. Since the 8

effect of the ionic strength consists mainly in the increased aggregation of bacterial cells (reduction 9

of the repulsive forces) that are then strained in the smallest pores, the salinity increase may not be 10

sufficient to modify the overall cell filtration rate if cell transport occurs mainly in the soil’s widest 11

pores. In other words, E. coli aggregation may occur over 0.1 M ionic strength, but does not induce 12

the formation of cells aggregates bigger than the medium size of the macropores that control water 13

percolation, as it may be observed in model porous media, for which cell attachment efficiency is 14

often controlled by solution ionic strength. 15

These results are in agreement with a recent study that compares E. coli cells migration in model 16

quartz sand and natural agricultural soil repacked in columns. 25 Although their agricultural soil 17

columns were not extracted as undisturbed cores (i.e. with preserved soil structure), Schinner et al. 18

25 observed an apparent attachment efficiency of E. coli that was lower in an agricultural-soil than 19

in pure quartz sand. The authors suggested that cell transport predictions based on attachment 20

efficiencies measured in sand experiments might underestimate bacterial migration potential in 21

agricultural soils (where surface properties are heterogeneous). This effect would be even more 22

pronounced in undisturbed soil columns (our case study), for which, pore network heterogeneity 23

and temporal evolution (structure and porosity of the soil) are also important factors to consider in 24

biotic or abiotic transport studies. 59 Similarly, Mishurov et al. 42 observed size exclusion effects in 25

16

a heterogeneous partially saturated sand column that led to preferential breakthrough of colloids as 1

compared to a water tracer. 2

3

3.6 Water flow effect 4

The effect of the flow velocity on bacterial cell transport was evaluated with two rainfall intensities 5

using bacterial suspensions of the same ionic strength (Table 2, B#5 and B#7). The breakthrough 6

curves of E. coli cells (Figure 7) observed at 10 and 20 mm h-1 appeared not significantly different 7

and presented a very similar shape. These results indicate that in our experimental conditions, the 8

water flow had no significant effect on the transport of bacteria in undisturbed soils (considering 9

the precision of the bacteria enumeration results). This result is in disagreement with previous 10

studies conducted with artificial sandy porous media. 1,9,29,60 In our study, the range of water flow 11

variation is rather narrow and the transport mode in the macropores is probably very different of 12

the flow regime prevalent in homogeneous sand columns, in which flow velocity effects have been 13

observed. These two points may explain the results observed in our study and the weak sensitivity 14

of E. coli cells transport to flow variations in an undisturbed soil core. Indeed, if bacteria are 15

naturally transported fast in the macropore network, doubling the water flow velocity in the 16

macropores might not significantly affect the mechanisms and kinetics of the reactive transport of 17

bacteria in the undisturbed soils. 18

19

4. Conclusion 20

In the present study we investigated the transport of Escherichia coli cells, used as surrogates to 21

pathogenic bacteria, through undisturbed cores of a natural brown leached soil. Our results 22

evidenced the importance of the structure of the pore network (and of its variation) of the 23

undisturbed soil for the movement of bacterial cells repeatedly injected in the cores. The temporal 24

evolution of the pore network (evidenced by Mercury Intrusion Porosimetry) was shown to modify 25

the main hydrodynamic process that evolved from a dominant matrix regime (first infiltrations) to 26

17

a dominant macroporosity transport regime of water and associated bacteria and solutes (last 1

infiltrations). These results stress the difficulty of working with undisturbed soil columns 2

compared to repacked model porous media. In contrast to model porous media, for which the 3

structure is stable and rather homogeneous, natural porous media are known to undergo structural 4

changes over time, mainly due to wetting/drying cycles or freezing. The main consequence of the 5

changes evidenced in this study was a huge acceleration of the transport of E. coli cells repeatedly 6

injected in the undisturbed soil cores. We also reported a relative insensitivity of bacteria transport 7

to factors previously shown to control bacterial transport in homogeneous repacked porous media 8

(water flow and ionic strength), in relation with the fast transport of bacteria in the macropores. 9

Our study points out in particular the importance of using undisturbed natural porous media to 10

evaluate the mobility of bacteria in the field, since soil hydrodynamic properties appeared 11

dominating all the other physical chemical factors. In this way, these results improve our 12

understanding of bio-colloid mobilization in undisturbed natural soils, stressing in particular the 13

need to account for preferential flow in bacteria or colloid transport modelling. 14

15

5. Acknowledgements 16

The authors gratefully acknowledge the technical plateau MOME of the GIS Envirhonalp from the 17

French Rhône-Alpes region, located at LTHE Grenoble, for the access to its microbiology 18

facilities. 19

20

6. References 21

1. C.P. Gerba and G. Bitton. Microbial pollutants: their survival and transport pattern to 22

groundwater. In: Groundwater pollution Microbiology. Eds G. Bitton and C.P. Gerba. 23

377p. John Wiley, New York. 1984. 24

2. J.T. Gannon, V.B. Manilal, M. Alexander, Appl. Environ. Microbiol., 1991, 57, 190–193. 25

3. W.P. Johnson, B.E. Logan, Water Res, 1996, 30, 923–931. 26

18

4. N. Yee, J.B. Fein, C.J. Daughney, Geochim. Cosmochim. Acta, 2000, 64, 609-617. 1

5. V. Guiné, Martins, J. M. F., Gaudet, J. P., J. Phys. IV, 2003, 107, 593-596. 2

6. LJ. Monrozier, P. Guez, A. Chalamet, R. Bardin, J. Martins and J.P. Gaudet. Sci. Total 3

Environ.,1993, 36, 121-133. 4

7. M. Ledin, Earth Sci. Rev., 2000, 51, 1-31. 5

8. L. Pang, Close, M. E., Noonan, M. J., Flintoft, M. J., van den Brink, P. A., J. Environ. 6

Qual., 2005, 34, 237-247. 7

9. J. F. Schivjen, S. M. Hassanizadeh and H. A. M. de Bruin, J. Contam. Hydrol. 2002, 58, 8

243-259. 9

10. T. R. Ginn, Wood, B. D., Nelson, K. E., Scheibe, T. D., Murphy, E. M., Clement, T. P., 10

Adv. Water Resour., 2002, 25, 1017-1042. 11

11. F. Celico, M. Varcamonti, M. Guida and G. Naclerio, Appl. Environ. Microbiol. 2004, 60, 12

2843-2847. 13

12. G. Naclerio, G. Fardella, G. Marzullo and F. Celico, J. Colloids Surf. B, 2009, 70, 25-28. 14

13. T.B. Bech, K. Johnsen, A. Dalsgaard, M. Laegdsmand, O.H. Jacobsen, and C.S. Jacobsen, 15

Appl. Environ. Microbiol. 2010, 76,710-714. 16

14. G. Gargiulo, S. Bradford, J. Simunek, P. Ustohal, H. Vereecken and E. Klumpp, J. Contam. 17

Hydrol, 2007, 92, 255–273. 18

15. J. W. A. Foppen, J. Hydrol. 2002, 263, 198-216. 19

16. A.A. Bordalo and J. Savva-Bordalo, Water Res. 2007, 41, 2978-2986. 20

17. J. Leber, M.M. Rahman, K.M. Ahmed, B. Mailloux, A. van Geen, Ground Water, 2011, 49, 21

111-123. 22

18. R.M. Kretchmar, D. Borkovek, Grolimund and M. Elimelech, Adv. Agron., 1999, 66, 121-23

193. 24

19. L. W. de Jonge, C. Kjaergaard, and P. Moldrup, Vadose Zone J., 2004, 3, 321-325. 25

19

20. W. Zhang, V.L. Morales, M.E. Cakmak, A.E. Salvucci, L.D. Geohring, A.G. Hay, J.-Y. 1

Parlange and T.S. Steenhuis, Environ. Sci. Technol., 2010, 44, 4965-4972. 2

21. D. Ongeng, C. Muyanja, A.H. Geeraerd, D. Springael and J. Ryckeboer, J. Appl. 3

Microbiol., 2011, 110, 1007–1022. 4

22. S. Rogers and J. Haines. Detecting and mitigating the environmental impact of fecal 5

pathogens originating from confined animal feeding operations, review. EPA report 6

EPA/600/R-06/021. 2005. 7

23. J.M. Passmore, D.L. Rudolph, M.M.F. Mesquita, E.E. Cey and M.B. Emelko, Water Res., 8

2010, 44, 1235–1245. 9

24. H.-H. Yang, Morrow, J. B., Grasso, D., Vinopal, R. T., Smets, B. F. Environ. Sci. Technol. 10

2006, 40, 6976-6982. 11

25. T. Schinner, Letzner, A., Liedtke, S., Castro, F. D., Eydelnant, I. A., Tufenkji, N. Water 12

Res. 2009, 44, 1182-1192. 13

26. T.G. Poulsen, Moldrup P., de Jonge, L.W., Komatsu T. Vadose Zone J., 2006, 5, 649-656. 14

27. N. Tufenkji, Adv. Water Resour., 2007, 30, 1455-1469. 15

28. G. Chen and K. A. Strevett, Environ. Microbiol. 2001, 3, 237-245. 16

29. S. C. Bradford, J. Simunek, S. L. Walker, Water Resour. Res. 2005, 42, W12S12, 17

doi,10.1029/2005WR004805. 18

30. J. W. Foppen, Van Herwerden, M., Schijven, J. F. Water Res., 2007, 41, 1743-1753. 19

31. G. Lutterodt, M. Basnet, J. W. A. Foppen, S. Uhlenbrook, Water Res., 2009, 43, 595-604. 20

32. J. Goldshmid, Zohar, D., Argamon, Y., Kott, Y. In Effects of dissolved salts on the 21

filtration of coliform bacteria in sand dunes, Jenkins S. H., Ed., Advances in Water 22

Pollution Research. Pergamon Press, New York, 1973, pp 147-155. 23

33. J. N. Ryan and M. Elimelech, J. Colloids Surf. A. 1996, 107, 1–56. 24

34. D. K. Powelson and A. L. Mills, J. Environ. Qual. 2001, 30, 238-245. 25

20

35. J. A. Redman, S. L. Walker and M. Elimelech, Environ. Sci. Technol. 2004, 38, 1777-1785. 1

36. T.K. Sen and K.C. Khilar, Adv. Coll. Inter. Sci. 2006, 19, 71-96. 2

37. M. Laegdsmand, K.G. Villholth, M. Ullum and Jensen K.H. Geoderma, 1999, 93, 33-59. 3

38. V. Lazouskaya, Y. Jin, and D. Or, J. Colloid Interf. Sci., 2006, 303, 171-184. 4

39. H. Dong, T. C. Onstott, M. D. DeFlaun, M. E. Fuller, S. H. Scheibe, R. K. Rothmel and 5

B.J. Mailloux, Environ. Sci. Technol., 2002, 36, 891-900. 6

40. S.A. Bradford and S. Torkzaban. Vadose Zone J., 2008, 7, 667-681. 7

41. S. Majdalani, E. Michel, L. Di Pietro, R. Angulo-Jaramillo and M. Rousseau. Water 8

Resour. Res., 2007, 43. W05401 DOI, 10.1029/2006WR005057 9

42. M. Mishurov, A. Yakirevich and N. Weisbrod, Environ. Sci. Technol., 2008, 42, 1066-1071 10

43. B.A. Macler, J. C. Merkle. Hydrogeol. J., 2000, 8, 29-40. 11

44. V. Guiné, L. Spadini, M. Muris, G. Sarret, C. Delolme, J. P. Gaudet and J. M. F. Martins, 12

Environ. Sci. Technol., 2006, 40, 1806-1813. 13

45. J. M. Martins, L.J. Monrozier, A. Chalamet, R. Bardin, Chemosphere, 1997, 35, 1637-1650 14

46. F. Baptist, L. Singer, J.C. Clement, C. Gallet, R. Guillemin, J. M. F. Martins, L. Sage, B. 15

Shavanaz, P. Choller and R. Geremia, Environ. Microbiol., 2008, 10, 799-809. 16

47. C. Duwig, B. Normand, M. Vauclin, G. Vachaud, S. R. Green and T. Becquer, Vadose 17

Zone J, 2003, 2, 76-89. 18

48. S. Majdalani, E. Michel, L. Di Pietro and R. Angulo-Jaramillo, Eur. J. Soil Sci., 2008, 59, 19

147-155. 20

49. J.M.F. Martins and A. Mermoud. Chemosphere, 1999, 38, 601-616. 21

50. V. Guiné, J. M. F. Martins, B. Causse, A. Durand and L. Spadini, Chem. Geol., 2007, 236, 22

266-280. 23

51. Z.G. Pei, X.Q. Shan, B. Wen, B. He, T. Liu, Y.N. Xie and G. Owens, Environ. Sci. 24

Technol., 2008, 42, 6849-6854. 25

21

52. I. Mubarak, J.C. Mailhol, R. Angulo-Jaramillo, P. Ruelle, P. Boivin, M. and Khaledian, 1

oderma, 2009, 150, 158-165. 2

53. J. Dorner, P. Sandoval and D. Dec. J. Soil Sci. & Plant Nutri., 2010, 10, 495-508. 3

54. K. Price, C.R. Jackson and A.J. Parkera, J. Hydrol., 2010, 383, 256-268. 4

55. H. Soulia, J.-M. Fleureaua and M. Trabelsi Ayadib. Geoderma, 2008, 145, 1–7. 5

56. M. Besnard, M. Burkhardt, R. Kasteel, J. Vanderborght and H. Vereecken, Eur. J. Soil Sci. 6

2008, 59, 82–93. 7

57. N. Toride, F.J. Leij and M.T. Van Genuchten. The CXTFIT Code for Estimating Transport 8

Parameters form Laboratory or Field Tracer Experiments Version 2.1. U.S. Salinity 9

Laboratory, Riverside, CA. 1999. 10

58. J.P. Gaudet, H. Jégat, G. Vachaud and P. Wierenga, Soil Sci. Soc. Am. J., 1977, 41, 665-71 11

59. C. Hinz, A. J. C. Sinke, J.M. Martins and H. Fluehler. J. Contam. Hydrol., 1998, 33, 1-3. 12

60. T. A. Camesano and B. E. Logan, Environ. Sci. Technol. 1998, 32, 1699-1708. 13

22

Table 1: Main physical chemical properties of the LCSA soil. 1 2

3 4 5 6 7 8 9 10

11 Table 2: Experimental conditions applied to the undisturbed soils cores A and B. 12 13

Experiment Number

Input bacteria concentration

(Cells mL-1)

Incoming solution ionic strength

(M)

Rainfall intensity (mm h-1)

A #1 1.02 109 DIW* 20 A #2 1.02 109 DIW 20 A #3 1.05 109 DIW 20 A #4 1.01 109 DIW 20 A #5 9.99 108 DIW 20 A #6 9.93 108 DIW 20 B # 1 1.00 109 DIW 20 B # 2 9.99 108 DIW 20 B # 3 1.01 109 DIW 20 B # 4 1.01 109 DIW 20 B # 5 9.50 108 8 10-3 20 B # 6 1.04 109 10-1 20 B # 7 1.00 109 8 10-3 10

* De-Ionized Water (10-7 M) 14 15 16

Clay (%)

Silt (%)

Sand (%)

CEC

(mEq Kg-1)

OC (%)

pHW

C/N

d50 (µm) Coefficient of uniformity (d60/d10)

Before infiltrations

After infiltrations

Before infiltrations

After infiltrations

LCSA Soil

21

48

31

114

1.05

5.8

8.7

136 ± 2.1

119 ± 1.9

12 ± 0.3

15 ± 0.3

23

1 2 Table 3. Zeta potentials of Escherichia coli cells and LCSA soil, and calculated DLVO 3 interaction parameters. Φmax is the height of the energy barrier and Φmin2 is the depth of the 4 second minimum of energy. 5 6

Electrolyte

Zeta potential (mV) (+/- 1STD) Energy

barrier Φmax/kBT

Secondary minimum

depth Φmin2/kBT

Secondary minimum separation

(nm) Escherichia coli

cells LCSA soil

De-ionized water (DIW)

-57.2 (+/-6) -20.1 (+/-0.9) 486 -4.27 10-5

5900

KBr

8 10-3

M -19.7 (+/-09) -14.2 (+/-) 72.0 -2.19

19.9

KBr

0.1 M -6.2 (+/-0.5) -6.0 (+/-6) - -

-

7

8

9

10

11

Table 4: Hydrodynamic parameters of the undisturbed soil columns determined with the 12 CXTFIT code 57 by fitting the bromide tracing experiments B#5 and B#6. 13

Parameter MR (%)

Pulse (-)

R (-)

ββββ (-)

ωωωω (-)

D (cm2 h-1)

αααα (h-1)

Tracer

101 ±2

0.44

1

0.14

0.69

3.4

0.056

MR is the average mass recovery, Pulse is the dimensionless pulse duration (relative to the 14 pore water volume), R is the retardation factor (fixed value), β is the mobile to total water 15 ratio, ω is the dimensionless mobile-immobile exchange coefficient, D is the dispersion 16 coefficient and α is the dimensional mobile-immobile exchange parameter. 17

18

24

1

2

Figure 1 3

Scheme of the experimental setup. The undisturbed soil column is 15 cm in diameter and 25 4

cm in length. 5

6

7

Effluent

Pump

Solute tank

Rainfall simulator

Undisturbed soil core

Fraction collector

Needles

25

Figure 2 1

Rel

ativ

e C

ell c

once

ntra

tion

(C/C

o)

0

0.2

0.4

0.6

0.8

1

0 1 2

Rel

ativ

e C

ell c

once

ntra

tion

(C/C

o)

0

0.2

0.4

0.6

0.8

1

0 1 2

0

0.2

0.4

0.6

0.8

1

0 1 2 0

0.2

0.4

0.6

0.8

1

0 1 2

0

0.2

0.4

0.6

0.8

1

0 1 2 0

0.2

0.4

0.6

0.8

1

0 1 2

0

0.2

0.4

0.6

0.8

1

0 1 2

0

0.2

0.4

0.6

0.8

1

0 1 2

0

0.2

0.4

0.6

0.8

1

0 1 2

V/Vo

Breakthrough curves of E. coli cells repeatedly injected in two undisturbed cores (A and B) of the LCSA soil. 800 mL of E. coli cells (109 cells mL) suspension were injected 6 times (column A) and 4 times (column B) consecutively with a draining period of 2 days between each experiment

0

0.2

0.4

0.6

0.8

1

0 1 2

to ensure identical initial conditions with regard to the water content and distribution within the soil. Rainfall intensity was 20 mm h-1 and the ionic strength was 10-7 M (deionised water). The error on cell concentration measurements is about ±10 % (measured in triplicates). (Dashed line: E. coli injection, Diamonds: E. coli break-through curve).

V/Vo

2

A#5 MB: 90%

A#3 MB: 90%

A#1 MB: 90%

A#2 MB: 91%

A#4 MB: 107%

B#1 MB : 92%

B#2 MB: 90%

B#3 MB: 109%

B#4 MB: 103%

A#6 MB: 92%

26

0

0,25

0,5

0,75

1

0 0,5 1 1,5

V/Vo

C/C

o

Figure 3

Comparison of E. coli cells first (black symbols) and last (opened symbols) breakthrough curves in the soil cores A (squares) and B (circles).

27

0,0

1,0

2,0

3,0

4,0

5,0

Par

ticle

cla

ss v

olum

e (%

)

A

1

0

0,2

0,4

0,6

0,8

1

0,01 0,1 1 10 100 1000 10000

Particle or pore size (µm)

Rel

ativ

e cu

mul

ated

por

e vo

lum

e .

Before bacteria infiltrationsAfter bacteria infiltrations

B

2 Figure 4 : A/ Comparison of the Particle Size Distributions (PSD) of the LCSA soil before 3

and after the repeated bacteria infiltrations. The measurements were performed in triplicate on 4

a few grams of the topsoil of core A with a MasterSizer 2000 laser granulometer (Malvern 5

Instruments). B/ Comparison of the pore size distribution of aggregates of surface soil 6

collected in core A, before and after the repeated bacterial suspensions infiltrations. 7

Measurements were performed by Mercury Intrusion Porosimetry (MIP) (Autopore III 9420, 8

Micromeritics) in triplicate with air-dried soil macro-aggregates of about 1 cm3.9

28

0

0,25

0,5

0,75

1

0 0,5 1 1,5 2

V/Vo

Rel

ativ

e B

rom

ide

conc

entr

atio

n (C

/Co)

Tracer Fitted

Bromide MB = 101% +/- 2

1

Figure 5 2

29

Breakthrough curve of bromide tracer in the undisturbed core B of LCSA soil performed at 20 mm h-1 (Exp. B#5 and B#6). Error bars 1

correspond to one standard deviation calculated with duplicates. Solid line corresponds to the tracer breakthrough curve fitted with CXTFIT 57 2

0

0,25

0,5

0,75

1

0 0,5 1 1,5 2

V/V0

C/C

0

Pulse B#4 I = 10-7M, MB = 109% B#5 I = 8.10-3M, MB = 103% B#6 I = 10-1M, MB = 90% Tracer fitted

3

Figure 6 4

30

Effect of the solution ionic strength on the transport of Escherichia coli cells in an undisturbed core of the LCSA soil. The rainfall intensity was 1

20 mm h-1. Symbols: relative E. coli concentrations measured in experiments B#4 to B#6. Solid line: fitted tracer breakthrough curve. 2

31

0

0,25

0,5

0,75

1

0 0,5 1 1,5 2

V/Vo

Rel

ativ

e ce

ll co

ncen

tratio

n (C

/Co)

PulseB#5: R = 20 mm h-1 (MB: 90%)B#7: R = 10 mm h-1 (MB: 98%)

1

Figure 7 2

Effect of the water flow velocity on the transport of E. coli cells in an undisturbed core of the LCSA soil (Exp. B#5 and B#7). The ionic strength 3

of the bacterial suspension was set to 8.10-3 M with KBr. 4